con web ui

- Fix get_partitions_from_year(): RAWDATACOR end_year now 2030 (was 2024)
  - RAWDATACOR has partitions d12-d17 for years 2025-2030, not just part0-part10
- Update year_to_partition_name() for RAWDATACOR: handle both part and d suffix
  - Years 2014-2024: use "part" suffix with formula (year - 2014)
  - Years 2025-2030: use "d" suffix with formula (year - 2013) for d12-d17
- Clamp year to range [2014, 2030] for RAWDATACOR
- Update docstring examples to reflect new mapping behavior
- Now correctly generates partitions like: part8, part9, part10, d12, d13, ..., d17

Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>

.env.example

@@ -13,10 +13,10 @@ POSTGRES_PASSWORD=your_postgres_password
 POSTGRES_DATABASE=migrated_db
 
 # Migration Settings
-BATCH_SIZE=10000
 LOG_LEVEL=INFO
 DRY_RUN=false
-CONSOLIDATION_GROUP_LIMIT=50000
+CONSOLIDATION_GROUP_LIMIT=40000
+PROGRESS_LOG_INTERVAL=10000
 
 # Performance Testing
 BENCHMARK_OUTPUT_DIR=benchmark_results

CHANGELOG.md

@@ -0,0 +1,121 @@
+# Changelog
+
+## [Current] - 2025-12-30
+
+### Added
+- **Consolidation-based incremental migration**: Uses consolidation keys `(UnitName, ToolNameID, EventDate, EventTime)` instead of timestamps
+- **MySQL ID optimization**: Uses `MAX(mysql_max_id)` from PostgreSQL to filter MySQL queries, avoiding full table scans
+- **State management in PostgreSQL**: Replaced JSON file with `migration_state` table for more reliable tracking
+- **Sync utility**: Added `scripts/sync_migration_state.py` to sync state with actual data
+- **Performance optimization**: MySQL queries now instant using PRIMARY KEY filter
+- **Data quality validation**: Automatically validates and logs invalid consolidation keys to dedicated error files
+- **Error logging**: Invalid keys (null dates, empty tool IDs, corrupted Java strings) are logged and skipped during migration
+- **Better documentation**: Consolidated and updated all documentation files
+
+### Changed
+- **Incremental migration**: Now uses consolidation keys instead of timestamp-based approach
+- **Full migration**: Improved to save global `last_key` after completing all partitions
+- **State tracking**: Moved from `migration_state.json` to PostgreSQL table `migration_state`
+- **Query performance**: Added `min_mysql_id` parameter to `fetch_consolidation_keys_after()` for optimization
+- **Configuration**: Renamed `BATCH_SIZE` to `CONSOLIDATION_GROUP_LIMIT` to better reflect what it controls
+- **Configuration**: Added `PROGRESS_LOG_INTERVAL` to control logging frequency
+- **Configuration**: Added `BENCHMARK_OUTPUT_DIR` to specify benchmark results directory
+- **Documentation**: Updated README.md, MIGRATION_WORKFLOW.md, QUICKSTART.md, EXAMPLE_WORKFLOW.md with current implementation
+- **Documentation**: Corrected index and partitioning documentation to reflect actual PostgreSQL schema:
+  - Uses `event_timestamp` (not separate event_date/event_time)
+  - Primary key includes `event_year` for partitioning
+  - Consolidation key is UNIQUE (unit_name, tool_name_id, event_timestamp, event_year)
+
+### Removed
+- **migration_state.json**: Replaced by PostgreSQL table
+- **Timestamp-based migration**: Replaced by consolidation key-based approach
+- **ID-based resumable migration**: Consolidated into single consolidation-based approach
+- **Temporary debug scripts**: Cleaned up all `/tmp/` debug files
+
+### Fixed
+- **Incremental migration performance**: MySQL queries now ~1000x faster with ID filter
+- **State synchronization**: Can now sync `migration_state` with actual data using utility script
+- **Duplicate handling**: Uses `ON CONFLICT DO NOTHING` to prevent duplicates
+- **Last key tracking**: Properly updates global state after full migration
+- **Corrupted data handling**: Both full and incremental migrations now validate keys and log errors instead of crashing
+
+### Error Logging
+
+Both full and incremental migrations now handle corrupted consolidation keys gracefully:
+
+**Error files:**
+- Full migration: `migration_errors_<table>_<partition>.log` (e.g., `migration_errors_rawdatacor_p2024.log`)
+- Incremental migration: `migration_errors_<table>_incremental_<timestamp>.log` (e.g., `migration_errors_rawdatacor_incremental_20260101_194500.log`)
+
+Each incremental migration creates a new timestamped file to preserve error history across runs.
+
+**File format:**
+```
+# Migration errors for <table> partition <partition>
+# Format: UnitName|ToolNameID|EventDate|EventTime|Reason
+
+ID0350||0000-00-00|0:00:00|EventDate is invalid: 0000-00-00
+[Ljava.lang.String;@abc123|TOOL1|2024-01-01|10:00:00|UnitName is corrupted Java string: [Ljava.lang.String;@abc123
+UNIT1||2024-01-01|10:00:00|ToolNameID is NULL or empty
+```
+
+**Behavior:**
+- Invalid keys are automatically skipped to prevent migration failure
+- Each skipped key is logged with the reason for rejection
+- Total count of skipped keys is reported at the end of migration
+- Empty error files (no errors) are automatically deleted
+
+### Migration Guide (from old to new)
+
+If you have an existing installation with `migration_state.json`:
+
+1. **Backup your data** (optional but recommended):
+   ```bash
+   cp migration_state.json migration_state.json.backup
+   ```
+
+2. **Run full migration** to populate `migration_state` table:
+   ```bash
+   python main.py migrate full
+   ```
+
+3. **Sync state** (if you have existing data):
+   ```bash
+   python scripts/sync_migration_state.py
+   ```
+
+4. **Remove old state file**:
+   ```bash
+   rm migration_state.json
+   ```
+
+5. **Run incremental migration**:
+   ```bash
+   python main.py migrate incremental --dry-run
+   python main.py migrate incremental
+   ```
+
+### Performance Improvements
+
+- **MySQL query time**: From 60+ seconds to <0.1 seconds (600x faster)
+- **Consolidation efficiency**: Multiple MySQL rows → single PostgreSQL record
+- **State reliability**: PostgreSQL table instead of JSON file
+
+### Breaking Changes
+
+- `--state-file` parameter removed from incremental migration (no longer uses JSON)
+- `--use-id` flag removed (consolidation-based approach is now default)
+- Incremental migration requires full migration to be run first
+- `BATCH_SIZE` environment variable renamed to `CONSOLIDATION_GROUP_LIMIT` (update your .env file)
+
+## [Previous] - Before 2025-12-30
+
+### Features
+- Full migration support
+- Incremental migration with timestamp tracking
+- JSONB transformation
+- Partitioning by year
+- GIN indexes for JSONB queries
+- Benchmark system
+- Progress tracking
+- Rich logging

EXAMPLE_WORKFLOW.md

@@ -74,8 +74,9 @@ nano .env
 # POSTGRES_PASSWORD=password123
 # POSTGRES_DATABASE=production_migrated
 #
-# BATCH_SIZE=50000        # Large batches for speed
 # LOG_LEVEL=INFO
+# CONSOLIDATION_GROUP_LIMIT=80000    # Large batches for speed
+# PROGRESS_LOG_INTERVAL=20000
 ```
 
 ### 4. Verifica Configurazione (5 min)
@@ -335,8 +336,11 @@ python main.py setup --create-schema
 
 ### Migrazione molto lenta
 ```bash
-# Aumentare batch size temporaneamente
-# Editare .env: BATCH_SIZE=100000
+# Aumentare consolidation group limit temporaneamente
+# Editare .env: CONSOLIDATION_GROUP_LIMIT=100000
+
+# Ridurre logging
+# Editare .env: PROGRESS_LOG_INTERVAL=50000
 
 # Oppure verificare:
 # - Latency rete MySQL↔PostgreSQL

MIGRATION_WORKFLOW.md

@@ -2,244 +2,396 @@
 
 ## Overview
 
-This tool supports three migration modes:
+This tool implements **consolidation-based incremental migration** for two tables:
+- **RAWDATACOR**: Raw sensor measurements
+- **ELABDATADISP**: Elaborated/calculated data
 
-1. **Full Migration** (`full_migration.py`) - Initial complete migration
-2. **Incremental Migration (Timestamp-based)** - Sync changes since last migration
-3. **Incremental Migration (ID-based)** - Resumable migration from last checkpoint
+Both tables use **consolidation keys** to group and migrate data efficiently.
 
 ---
 
-## 1. Initial Full Migration
+## Migration Modes
 
-### First Time Setup
+### 1. Full Migration
+
+Initial migration of all historical data, migrating one partition at a time.
 
 ```bash
-# Create the PostgreSQL schema
-python main.py setup --create-schema
+# Migrate all partitions for all tables
+python main.py migrate full
 
-# Run full migration (one-time)
-python main.py migrate --full RAWDATACOR
-python main.py migrate --full ELABDATADISP
+# Migrate specific table
+python main.py migrate full --table RAWDATACOR
+
+# Migrate specific partition (year-based)
+python main.py migrate full --table ELABDATADISP --partition 2024
+
+# Dry-run to see what would be migrated
+python main.py migrate full --dry-run
 ```
 
-**When to use:** First time migrating data or need complete fresh migration.
-
 **Characteristics:**
-- Fetches ALL rows from MySQL
-- No checkpoint tracking
-- Cannot resume if interrupted
-- Good for initial data load
+- Migrates data partition by partition (year-based)
+- Uses consolidation groups for efficiency
+- Tracks progress in `migration_state` table (PostgreSQL)
+- Can resume from last completed partition if interrupted
+- Uses `mysql_max_id` optimization for performance
 
 ---
 
-## 2. Timestamp-based Incremental Migration
+### 2. Incremental Migration
 
-### For Continuous Sync (Recommended for most cases)
+Sync only new data since the last migration.
 
 ```bash
-# After initial full migration, use incremental with timestamps
-python main.py migrate --incremental RAWDATACOR
-python main.py migrate --incremental ELABDATADISP
+# Migrate new data for all tables
+python main.py migrate incremental
+
+# Migrate specific table
+python main.py migrate incremental --table ELABDATADISP
+
+# Dry-run to check what would be migrated
+python main.py migrate incremental --dry-run
 ```
 
-**When to use:** Continuous sync of new/updated records.
-
 **Characteristics:**
-- Tracks `created_at` (RAWDATACOR) or `updated_at` (ELABDATADISP)
-- Uses JSON state file (`migration_state.json`)
-- Only fetches rows modified since last run
-- Perfect for scheduled jobs (cron, airflow, etc.)
-- Syncs changes but NOT deletions
+- Uses **consolidation keys** to identify new records:
+  - `(UnitName, ToolNameID, EventDate, EventTime)`
+- Tracks last migrated key in `migration_state` table
+- Optimized with `min_mysql_id` filter for performance
+- Handles duplicates with `ON CONFLICT DO NOTHING`
+- Perfect for scheduled jobs (cron, systemd timers)
 
 **How it works:**
-1. First run: Returns with message "No previous migration found" - must run full migration first
-2. Subsequent runs: Only fetches rows where `created_at` > last_migration_timestamp
-3. Updates state file with new timestamp for next run
-
-**Example workflow:**
-```bash
-# Day 1: Initial full migration
-python main.py migrate --full RAWDATACOR
-
-# Day 1: Then incremental (will find nothing new)
-python main.py migrate --incremental RAWDATACOR
-
-# Day 2, 3, 4: Daily syncs via cron
-python main.py migrate --incremental RAWDATACOR
-```
+1. Retrieves `last_key` from `migration_state` table
+2. Gets `MAX(mysql_max_id)` from PostgreSQL table for optimization
+3. Queries MySQL: `WHERE id > max_mysql_id AND (key_tuple) > last_key`
+4. Migrates new consolidation groups
+5. Updates `migration_state` with new `last_key`
 
 ---
 
-## 3. ID-based Incremental Migration (Resumable)
+## Consolidation Keys
 
-### For Large Datasets or Unreliable Connections
+Both tables use consolidation to group multiple measurements into a single JSONB record.
 
-```bash
-# First run
-python main.py migrate --incremental RAWDATACOR --use-id
+### Consolidation Key Structure
 
-# Can interrupt and resume multiple times
-python main.py migrate --incremental RAWDATACOR --use-id
+```sql
+(UnitName, ToolNameID, EventDate, EventTime)
 ```
 
-**When to use:**
-- Large datasets that may timeout
-- Need to resume from exact last position
-- Network is unstable
+### Data Quality Validation
 
-**Characteristics:**
-- Tracks `last_id` instead of timestamp
-- Updates state file after EACH BATCH (not just at end)
-- Can interrupt and resume dozens of times
-- Resumes from exact record ID where it stopped
-- Works with `migration_state.json`
+The migration automatically validates and logs invalid consolidation keys:
+- `EventDate IS NULL` or `EventDate = '0000-00-00'`
+- `ToolNameID IS NULL` or `ToolNameID = ''` (empty string)
+- `UnitName IS NULL` or `UnitName = ''` (empty string)
+- `UnitName` starting with `[L` (corrupted Java strings like `[Ljava.lang.String;@...`)
+- `EventTime IS NULL`
 
-**How it works:**
-1. First run: Starts from beginning (ID = 0)
-2. Each batch: Updates state file with max ID from batch
-3. Interrupt: Can stop at any time
-4. Resume: Next run continues from last ID stored
-5. Continues until all rows processed
+Invalid keys are:
+- **Logged to error files** for tracking and analysis
+- **Skipped automatically** to prevent migration failures
+- **Counted and reported** at the end of migration
 
-**Example workflow for large dataset:**
-```bash
-# Start ID-based migration (will migrate in batches)
-python main.py migrate --incremental RAWDATACOR --use-id
+Error log files:
+- Full migration: `migration_errors_<table>_<partition>.log` (e.g., `migration_errors_rawdatacor_p2024.log`)
+- Incremental migration: `migration_errors_<table>_incremental_<timestamp>.log` (e.g., `migration_errors_rawdatacor_incremental_20260101_194500.log`)
 
-# [If interrupted after 1M rows processed]
+Each incremental migration creates a new timestamped file to preserve history.
 
-# Resume from ID 1M (automatically detects last position)
-python main.py migrate --incremental RAWDATACOR --use-id
+### Why Consolidation?
 
-# [Continues until complete]
+Instead of migrating individual sensor readings, we:
+1. **Group** all measurements for the same (unit, tool, date, time)
+2. **Transform** 16-25 columns into structured JSONB
+3. **Migrate** as a single consolidated record
+
+**Example:**
+
+MySQL has 10 rows for `(Unit1, Tool1, 2024-01-01, 10:00:00)`:
+```
+id  | UnitName | ToolNameID | EventDate  | EventTime | Val0  | Val1  | ...
+1   | Unit1    | Tool1      | 2024-01-01 | 10:00:00  | 23.5  | 45.2  | ...
+2   | Unit1    | Tool1      | 2024-01-01 | 10:00:00  | 23.6  | 45.3  | ...
+...
+```
+
+PostgreSQL gets 1 consolidated record:
+```json
+{
+  "unit_name": "Unit1",
+  "tool_name_id": "Tool1",
+  "event_timestamp": "2024-01-01 10:00:00",
+  "measurements": {
+    "0": {"value": 23.5, "unit": "°C"},
+    "1": {"value": 45.2, "unit": "bar"},
+    ...
+  },
+  "mysql_max_id": 10
+}
 ```
 
 ---
 
 ## State Management
 
-### State File Location
-```
-migration_state.json  # In project root
+### Migration State Table
+
+The `migration_state` table in PostgreSQL tracks migration progress:
+
+```sql
+CREATE TABLE migration_state (
+    table_name VARCHAR(50),
+    partition_name VARCHAR(50),
+    last_key JSONB,  -- Last migrated consolidation key
+    started_at TIMESTAMP,
+    completed_at TIMESTAMP,
+    total_rows INTEGER,
+    status VARCHAR(20)
+);
 ```
 
-### State File Content (Timestamp-based)
-```json
-{
-  "rawdatacor": {
-    "last_timestamp": "2024-12-11T19:30:45.123456",
-    "last_updated": "2024-12-11T19:30:45.123456",
-    "total_migrated": 50000
-  }
-}
+### State Records
+
+- **Per-partition state**: Tracks each partition's progress
+  - Example: `('rawdatacor', '2024', {...}, '2024-01-15 10:30:00', 'completed', 1000000)`
+
+- **Global state**: Tracks overall incremental migration position
+  - Example: `('rawdatacor', '_global', {...}, NULL, NULL, 0, 'in_progress')`
+
+### Checking State
+
+```sql
+-- View all migration state
+SELECT * FROM migration_state ORDER BY table_name, partition_name;
+
+-- View global state (for incremental migration)
+SELECT * FROM migration_state WHERE partition_name = '_global';
 ```
 
-### State File Content (ID-based)
-```json
-{
-  "rawdatacor": {
-    "last_id": 1000000,
-    "total_migrated": 1000000,
-    "last_updated": "2024-12-11T19:45:30.123456"
-  }
-}
+---
+
+## Performance Optimization
+
+### MySQL ID Filter
+
+The incremental migration uses `MAX(mysql_max_id)` from PostgreSQL to filter MySQL queries:
+
+```sql
+SELECT UnitName, ToolNameID, EventDate, EventTime
+FROM RAWDATACOR
+WHERE id > 267399536  -- max_mysql_id from PostgreSQL
+  AND (UnitName, ToolNameID, EventDate, EventTime) > (?, ?, ?, ?)
+GROUP BY UnitName, ToolNameID, EventDate, EventTime
+ORDER BY UnitName, ToolNameID, EventDate, EventTime
+LIMIT 10000
 ```
 
-### Reset Migration State
-```python
-from src.migrator.state import MigrationState
+**Why this is fast:**
+- Uses PRIMARY KEY index on `id` to skip millions of already-migrated rows
+- Tuple comparison only applied to filtered subset
+- Avoids full table scans
 
-state = MigrationState()
+### Consolidation Group Batching
 
-# Reset specific table
-state.reset("rawdatacor")
-
-# Reset all tables
-state.reset()
-```
+Instead of fetching individual rows, we:
+1. Fetch 10,000 consolidation keys at a time
+2. For each key, fetch all matching rows from MySQL
+3. Transform and insert into PostgreSQL
+4. Update state every batch
 
 ---
 
 ## Recommended Workflow
 
-### For Daily Continuous Sync
-```bash
-# Week 1: Initial setup
-python main.py setup --create-schema
-python main.py migrate --full RAWDATACOR
-python main.py migrate --full ELABDATADISP
+### Initial Setup (One-time)
 
-# Week 2+: Daily incremental syncs (via cron job)
-# Schedule: `0 2 * * * cd /path/to/project && python main.py migrate --incremental RAWDATACOR`
-python main.py migrate --incremental RAWDATACOR
-python main.py migrate --incremental ELABDATADISP
+```bash
+# 1. Configure .env file
+cp .env.example .env
+nano .env
+
+# 2. Create PostgreSQL schema
+python main.py setup --create-schema
+
+# 3. Run full migration
+python main.py migrate full
 ```
 
-### For Large Initial Migration
-```bash
-# If dataset > 10 million rows
-python main.py setup --create-schema
-python main.py migrate --incremental RAWDATACOR --use-id  # Can interrupt/resume
+### Daily Incremental Sync
 
-# For subsequent syncs, use timestamp
-python main.py migrate --incremental RAWDATACOR  # Timestamp-based
+```bash
+# Run incremental migration (via cron or manual)
+python main.py migrate incremental
+```
+
+**Cron example** (daily at 2 AM):
+```cron
+0 2 * * * cd /path/to/mysql2postgres && python main.py migrate incremental >> /var/log/migration.log 2>&1
 ```
 
 ---
 
-## Key Differences at a Glance
+## Resuming Interrupted Migrations
 
-| Feature | Full | Timestamp | ID-based |
-|---------|------|-----------|----------|
-| Initial setup | ✅ Required first | ✅ After full | ✅ After full |
-| Sync new/updated | ❌ No | ✅ Yes | ✅ Yes |
-| Resumable | ❌ No | ⚠️ Partial* | ✅ Full |
-| Batched state tracking | ❌ No | ❌ No | ✅ Yes |
-| Large datasets | ⚠️ Risky | ✅ Good | ✅ Best |
-| Scheduled jobs | ❌ No | ✅ Perfect | ⚠️ Unnecessary |
+### Full Migration
 
-*Timestamp mode can resume, but must wait for full batch to complete before continuing
+If interrupted, full migration resumes from the last completed partition:
+
+```bash
+# First run: migrates partitions 2014, 2015, 2016... (interrupted after 2020)
+python main.py migrate full --table RAWDATACOR
+
+# Resume: continues from partition 2021
+python main.py migrate full --table RAWDATACOR
+```
+
+### Incremental Migration
+
+Incremental migration uses the `last_key` from `migration_state`:
+
+```bash
+# Always safe to re-run - uses ON CONFLICT DO NOTHING
+python main.py migrate incremental
+```
 
 ---
 
-## Default Partitions
+## Syncing Migration State
 
-Both tables are partitioned by year (2014-2031) plus a DEFAULT partition:
-- **rawdatacor_2014** through **rawdatacor_2031** (yearly partitions)
-- **rawdatacor_default** (catches data outside 2014-2031)
+If `migration_state` becomes out of sync with actual data, use the sync utility:
 
-Same for ELABDATADISP. This ensures data with edge-case timestamps doesn't break migration.
+```bash
+# Sync migration_state with actual PostgreSQL data
+python scripts/sync_migration_state.py
+```
+
+This finds the most recent row (by `created_at`) and updates `migration_state._global`.
 
 ---
 
 ## Monitoring
 
-### Check Migration Progress
-```bash
-# View state file
-cat migration_state.json
+### Check Progress
 
-# Check PostgreSQL row counts
-psql -U postgres -h localhost -d your_db -c "SELECT COUNT(*) FROM rawdatacor;"
+```bash
+# View migration state
+psql -h localhost -U postgres -d migrated_db -c \
+  "SELECT table_name, partition_name, status, total_rows, completed_at
+   FROM migration_state
+   ORDER BY table_name, partition_name"
 ```
 
-### Common Issues
+### Verify Row Counts
 
-**"No previous migration found"** (Timestamp mode)
-- Solution: Run full migration first with `--full` flag
+```sql
+-- PostgreSQL
+SELECT COUNT(*) FROM rawdatacor;
+SELECT COUNT(*) FROM elabdatadisp;
 
-**"Duplicate key value violates unique constraint"**
-- Cause: Running full migration twice
-- Solution: Use timestamp-based incremental sync instead
+-- Compare with MySQL
+-- mysql> SELECT COUNT(DISTINCT UnitName, ToolNameID, EventDate, EventTime) FROM RAWDATACOR;
+```
 
-**"Timeout during migration"** (Large datasets)
-- Solution: Switch to ID-based resumable migration with `--use-id`
+---
+
+## Common Issues
+
+### "No previous migration found"
+
+**Cause**: Running incremental migration before full migration
+
+**Solution**: Run full migration first
+```bash
+python main.py migrate full
+```
+
+### "Duplicate key value violates unique constraint"
+
+**Cause**: Data already exists (shouldn't happen with ON CONFLICT DO NOTHING)
+
+**Solution**: Migration handles this automatically - check logs for details
+
+### Slow Incremental Migration
+
+**Cause**: `MAX(mysql_max_id)` query is slow (~60 seconds on large tables)
+
+**Solution**: This is expected and only happens once at start. The MySQL queries are instant thanks to the optimization.
+
+**Alternative**: Create an index on `mysql_max_id` in PostgreSQL (uses disk space):
+```sql
+CREATE INDEX idx_rawdatacor_mysql_max_id ON rawdatacor (mysql_max_id DESC);
+CREATE INDEX idx_elabdatadisp_mysql_max_id ON elabdatadisp (mysql_max_id DESC);
+```
+
+---
+
+## Key Technical Details
+
+### Tuple Comparison in MySQL
+
+MySQL supports lexicographic tuple comparison:
+
+```sql
+WHERE (UnitName, ToolNameID, EventDate, EventTime) > ('Unit1', 'Tool1', '2024-01-01', '10:00:00')
+```
+
+This is equivalent to:
+```sql
+WHERE UnitName > 'Unit1'
+   OR (UnitName = 'Unit1' AND ToolNameID > 'Tool1')
+   OR (UnitName = 'Unit1' AND ToolNameID = 'Tool1' AND EventDate > '2024-01-01')
+   OR (UnitName = 'Unit1' AND ToolNameID = 'Tool1' AND EventDate = '2024-01-01' AND EventTime > '10:00:00')
+```
+
+But much more efficient!
+
+### Partitioning in PostgreSQL
+
+Tables are partitioned by year (2014-2031):
+```sql
+CREATE TABLE rawdatacor_2024 PARTITION OF rawdatacor
+  FOR VALUES FROM (2024) TO (2025);
+```
+
+PostgreSQL automatically routes INSERTs to the correct partition based on `event_year`.
+
+### Indexes in PostgreSQL
+
+Both tables have these indexes automatically created:
+
+**Primary Key** (required for partitioned tables):
+```sql
+-- Must include partition key (event_year)
+UNIQUE (id, event_year)
+```
+
+**Consolidation Key** (prevents duplicates):
+```sql
+-- Ensures one record per consolidation group
+UNIQUE (unit_name, tool_name_id, event_timestamp, event_year)
+```
+
+**Query Optimization**:
+```sql
+-- Fast filtering by unit/tool
+(unit_name, tool_name_id)
+
+-- JSONB queries with GIN index
+GIN (measurements)
+```
+
+**Note**: All indexes are automatically created on all partitions when you run `setup --create-schema`.
 
 ---
 
 ## Summary
 
-- **Start with:** Full migration (`--full`) for initial data load
-- **Then use:** Timestamp-based incremental (`--incremental`) for daily syncs
-- **Switch to:** ID-based resumable (`--incremental --use-id`) if full migration is too large
+1. **Full migration**: One-time initial load, partition by partition
+2. **Incremental migration**: Daily sync of new data using consolidation keys
+3. **State tracking**: PostgreSQL `migration_state` table
+4. **Performance**: `mysql_max_id` filter + consolidation batching
+5. **Resumable**: Both modes can resume from interruptions
+6. **Safe**: ON CONFLICT DO NOTHING prevents duplicates

QUICKSTART.md

@@ -42,8 +42,9 @@ POSTGRES_USER=postgres
 POSTGRES_PASSWORD=pgpassword
 POSTGRES_DATABASE=migrated_db
 
-BATCH_SIZE=10000
 LOG_LEVEL=INFO
+CONSOLIDATION_GROUP_LIMIT=40000
+PROGRESS_LOG_INTERVAL=10000
 ```
 
 ### 3. Creare PostgreSQL in Incus
@@ -111,8 +112,11 @@ python main.py migrate full --table RAWDATACOR
 # Migrare solo i cambiamenti dal last sync
 python main.py migrate incremental
 
-# Con stato personalizzato
-python main.py migrate incremental --state-file daily_sync.json
+# Dry-run per vedere cosa verrebbe migrato
+python main.py migrate incremental --dry-run
+
+# Solo una tabella specifica
+python main.py migrate incremental --table RAWDATACOR
 ```
 
 ### Benchmark
@@ -221,8 +225,11 @@ python main.py setup --create-schema
 
 ### "Migration is slow"
 ```bash
-# Aumentare batch size in .env
-BATCH_SIZE=50000
+# Aumentare consolidation group limit in .env
+CONSOLIDATION_GROUP_LIMIT=80000
+
+# Oppure ridurre logging
+PROGRESS_LOG_INTERVAL=20000
 
 # Oppure ottimizzare MySQL
 mysql> FLUSH PRIVILEGES;

config.py

@@ -52,6 +52,7 @@ class MigrationSettings(BaseSettings):
     consolidation_group_limit: int = 10000
     log_level: str = "INFO"
     dry_run: bool = False
+    progress_log_interval: int = 50000
 
 
 class BenchmarkSettings(BaseSettings):
@@ -152,20 +153,32 @@ ELABDATADISP_FIELD_MAPPING = {
 # PostgreSQL Partition years (from both tables)
 PARTITION_YEARS = list(range(2014, 2032))  # 2014-2031
 
+# Consolidation key definition (same for both tables)
+# Multiple MySQL rows with same key but different NodeNum → 1 PostgreSQL row
+# MySQL source fields
+CONSOLIDATION_KEY_FIELDS = ["UnitName", "ToolNameID", "EventDate", "EventTime"]
+# Keys for tracking in migration_state.last_key (NOT actual PostgreSQL target columns)
+# Note: In PostgreSQL target, EventDate+EventTime become event_timestamp
+CONSOLIDATION_KEY_PG_FIELDS = ["unit_name", "tool_name_id", "event_date", "event_time"]
+
 # Table configurations - support both uppercase and lowercase keys
 _rawdatacor_config = {
     "mysql_table": "RAWDATACOR",
     "postgres_table": "rawdatacor",
-    "primary_key": "id",
-    "postgres_pk": "id",  # Primary key column name in PostgreSQL
-    "partition_key": "event_timestamp",
+    "mysql_pk": "id",  # MySQL primary key
+    "postgres_pk": "id",  # PostgreSQL auto-increment primary key
+    "mysql_max_id_field": "id",  # Field to track max ID from MySQL
+    "consolidation_key": CONSOLIDATION_KEY_FIELDS,
+    "consolidation_key_pg": CONSOLIDATION_KEY_PG_FIELDS,
 }
 _elabdatadisp_config = {
     "mysql_table": "ELABDATADISP",
     "postgres_table": "elabdatadisp",
-    "primary_key": "idElabData",
-    "postgres_pk": "id_elab_data",  # Primary key column name in PostgreSQL
-    "partition_key": "event_timestamp",
+    "mysql_pk": "idElabData",  # MySQL primary key
+    "postgres_pk": "id",  # PostgreSQL auto-increment primary key
+    "mysql_max_id_field": "idElabData",  # Field to track max ID from MySQL
+    "consolidation_key": CONSOLIDATION_KEY_FIELDS,
+    "consolidation_key_pg": CONSOLIDATION_KEY_PG_FIELDS,
 }
 
 TABLE_CONFIGS = {
@@ -174,3 +187,118 @@ TABLE_CONFIGS = {
     "elabdatadisp": _elabdatadisp_config,
     "ELABDATADISP": _elabdatadisp_config,
 }
+
+
+# Partition mapping utilities
+def year_to_partition_name(year: int, table: str) -> str:
+    """Map year to partition name.
+
+    Partition naming scheme (different for each table):
+    - RAWDATACOR: part0=2014, part1=2015, ..., part10=2024 (part{year-2014})
+    - ELABDATADISP: d0=2013, d1=2014, ..., d12=2025, ..., d17=2030 (d{year-2013})
+
+    Args:
+        year: Year (2013-2031, depending on table)
+        table: Table name (RAWDATACOR or ELABDATADISP)
+
+    Returns:
+        Partition name (e.g., "part8" for RAWDATACOR/2022, "d14" for ELABDATADISP/2026)
+
+    Raises:
+        ValueError: If year is out of range or table is unknown
+    """
+    table_upper = table.upper()
+
+    if table_upper == "RAWDATACOR":
+        # RAWDATACOR: 2014-2024 (part0-part10)
+        # RAWDATACOR: 2025-2030 (d12-d17)
+
+        if year < 2014:
+            year = 2014
+        elif year > 2030:
+            year = 2030
+        
+        if year < 2025:
+            suffix = "part"
+            d_year = 2014
+        else:
+            suffix = "d"
+            d_year = 2013  # Continue naming as d12, d13, ...
+
+        partition_index = year - d_year  # 2014→0, 2015→1, ..., 2024→10 - 2025→12, ..., 2030→17
+        return f"{suffix}{partition_index}"
+
+    elif table_upper == "ELABDATADISP":
+        # ELABDATADISP: 2013-2031 (d0-d18)
+        if year < 2013:
+            year = 2013
+        elif year > 2031:
+            year = 2031
+
+        partition_index = year - 2013  # 2013→0, 2014→1, ..., 2025→12, ..., 2031→18
+        return f"d{partition_index}"
+
+    else:
+        raise ValueError(f"Unknown table: {table}")
+
+
+def get_partitions_from_year(year: int, table: str) -> list[str]:
+    """Get list of partition names from a specific year onwards.
+
+    Args:
+        year: Starting year
+        table: Table name (RAWDATACOR or ELABDATADISP)
+
+    Returns:
+        List of partition names from that year to the latest available year
+
+    Example:
+        get_partitions_from_year(2022, "RAWDATACOR")
+        → ["part8", "part9", "part10", "d12", "d13", "d14", "d15", "d16", "d17"]  # 2022→part8, ..., 2024→part10, 2025→d12, ..., 2030→d17
+
+        get_partitions_from_year(2025, "ELABDATADISP")
+        → ["d12", "d13", "d14", "d15", "d16", "d17"]  # 2025-2030
+    """
+    table_upper = table.upper()
+    partitions = []
+
+    if table_upper == "RAWDATACOR":
+        end_year = 2030  # RAWDATACOR: part0-part10 (2014-2024) + d12-d17 (2025-2030)
+    elif table_upper == "ELABDATADISP":
+        end_year = 2030  # ELABDATADISP: d0-d17 (2013-2030)
+    else:
+        raise ValueError(f"Unknown table: {table}")
+
+    # Generate partitions for each year from start_year to end_year
+    for y in range(year, end_year + 1):
+        partition_name = year_to_partition_name(y, table)
+        # Avoid duplicates (can happen if mapping multiple years to same partition)
+        if not partitions or partitions[-1] != partition_name:
+            partitions.append(partition_name)
+
+    return partitions
+
+
+def date_string_to_partition_name(date_str: str, table: str) -> str:
+    """Extract year from date string and map to partition name.
+
+    Args:
+        date_str: Date string in format 'YYYY-MM-DD' (e.g., '2022-05-15')
+        table: Table name (RAWDATACOR or ELABDATADISP)
+
+    Returns:
+        Partition name (e.g., "part8" or "d8")
+
+    Example:
+        date_string_to_partition_name("2022-05-15", "RAWDATACOR") → "part8"
+    """
+    if not date_str or len(date_str) < 4:
+        # Default to 2014 if invalid date
+        return year_to_partition_name(2014, table)
+
+    try:
+        year = int(date_str[:4])
+        return year_to_partition_name(year, table)
+    except (ValueError, TypeError):
+        # Default to 2014 if can't parse
+        return year_to_partition_name(2014, table)

main.py

@@ -6,8 +6,9 @@ from pathlib import Path
 from config import get_settings
 from src.utils.logger import setup_logger, get_logger
 from src.transformers.schema_transformer import get_full_schema_script
-from src.migrator.full_migration import run_full_migration
-from src.migrator.incremental_migration import run_incremental_migration
+from src.migrator.full_migrator import run_full_migration
+from src.migrator.incremental_migrator import run_incremental_migration
+from src.migrator.parallel_migrator import run_parallel_migration
 from src.benchmark.performance_test import run_benchmark
 from src.connectors.postgres_connector import PostgreSQLConnector
 
@@ -30,7 +31,7 @@ def cli(ctx):
 )
 def setup(create_schema):
     """Setup PostgreSQL database."""
-    setup_logger(__name__)
+    setup_logger("")  # Configure root logger to show all module logs
 
     if not create_schema:
         click.echo("Usage: python main.py setup --create-schema")
@@ -74,18 +75,42 @@ def migrate():
     is_flag=True,
     help="Resume from last checkpoint if migration was interrupted"
 )
-def full(table, dry_run, resume):
+@click.option(
+    "--partition",
+    type=str,
+    default=None,
+    help="Only migrate this partition (for testing/debugging)"
+)
+@click.option(
+    "--parallel",
+    type=int,
+    default=None,
+    help="Number of parallel workers (e.g., --parallel 5 for 5 workers)"
+)
+def full(table, dry_run, resume, partition, parallel):
     """Perform full migration of all data."""
-    setup_logger(__name__)
+    setup_logger("")  # Configure root logger to show all module logs
 
     tables = ["RAWDATACOR", "ELABDATADISP"] if table == "all" else [table]
 
+    # Validate options
+    if parallel and partition:
+        click.echo("✗ Cannot use --parallel with --partition", err=True)
+        sys.exit(1)
+
     try:
         total_migrated = 0
 
         for tbl in tables:
-            click.echo(f"\nMigrating {tbl}...")
-            migrated = run_full_migration(tbl, dry_run=dry_run, resume=resume)
+            if parallel:
+                # Parallel migration mode
+                click.echo(f"\nMigrating {tbl} with {parallel} parallel workers...")
+                migrated = run_parallel_migration(tbl, num_workers=parallel, dry_run=dry_run, resume=resume)
+            else:
+                # Sequential migration mode
+                click.echo(f"\nMigrating {tbl}" + (f" (partition {partition})" if partition else "") + "...")
+                migrated = run_full_migration(tbl, dry_run=dry_run, resume=resume, partition=partition)
+
             total_migrated += migrated
             click.echo(f"✓ {tbl}: {migrated} rows migrated")
 
@@ -109,14 +134,9 @@ def full(table, dry_run, resume):
     is_flag=True,
     help="Show what would be done without modifying data"
 )
-@click.option(
-    "--state-file",
-    default="migration_state.json",
-    help="Path to migration state file"
-)
-def incremental(table, dry_run, state_file):
-    """Perform incremental migration since last sync."""
-    setup_logger(__name__)
+def incremental(table, dry_run):
+    """Perform incremental migration since last sync (based on consolidation keys)."""
+    setup_logger("")  # Set up root logger so all child loggers work
 
     tables = ["RAWDATACOR", "ELABDATADISP"] if table == "all" else [table]
 
@@ -125,7 +145,7 @@ def incremental(table, dry_run, state_file):
 
         for tbl in tables:
             click.echo(f"\nIncremental migration for {tbl}...")
-            migrated = run_incremental_migration(tbl, dry_run=dry_run, state_file=state_file)
+            migrated = run_incremental_migration(tbl, dry_run=dry_run)
             total_migrated += migrated
             if migrated > 0:
                 click.echo(f"✓ {tbl}: {migrated} rows migrated")
@@ -158,7 +178,7 @@ def incremental(table, dry_run, state_file):
 )
 def benchmark(iterations, output):
     """Run performance benchmarks comparing MySQL and PostgreSQL."""
-    setup_logger(__name__)
+    setup_logger("")  # Configure root logger to show all module logs
 
     try:
         click.echo("Running performance benchmarks...")
@@ -174,7 +194,7 @@ def benchmark(iterations, output):
 @cli.command()
 def info():
     """Show configuration information."""
-    setup_logger(__name__)
+    setup_logger("")  # Configure root logger to show all module logs
 
     settings = get_settings()
 
@@ -190,6 +210,7 @@ def info():
 
     click.echo("\n[Migration Settings]")
     click.echo(f"  Batch Size: {settings.migration.batch_size}")
+    click.echo(f"  Consolidation Group Limit: {settings.migration.consolidation_group_limit}")
     click.echo(f"  Log Level: {settings.migration.log_level}")
     click.echo(f"  Dry Run: {settings.migration.dry_run}")
 
@@ -198,5 +219,40 @@ def info():
     click.echo(f"  Iterations: {settings.benchmark.iterations}")
 
 
+@cli.command()
+@click.option(
+    "--port",
+    type=int,
+    default=7860,
+    help="Port to run the web interface on (default: 7860)"
+)
+@click.option(
+    "--share",
+    is_flag=True,
+    help="Create a public share link (useful for remote access)"
+)
+def web(port, share):
+    """Launch web-based GUI for migration monitoring and control."""
+    setup_logger("")  # Configure root logger to show all module logs
+
+    try:
+        from web_ui import launch_ui
+
+        click.echo(f"\n🚀 Starting Migration Dashboard on http://localhost:{port}")
+        if share:
+            click.echo("📡 Creating public share link...")
+
+        launch_ui(share=share, server_port=port)
+
+    except ImportError as e:
+        click.echo(f"✗ Failed to import web_ui module: {e}", err=True)
+        click.echo("Make sure gradio is installed: uv sync", err=True)
+        sys.exit(1)
+    except Exception as e:
+        logger.error(f"Web interface failed: {e}")
+        click.echo(f"✗ Web interface failed: {e}", err=True)
+        sys.exit(1)
+
+
 if __name__ == "__main__":
     cli(obj={})

pyproject.toml

@@ -13,4 +13,7 @@ dependencies = [
     "pydantic>=2.5.0",
     "pydantic-settings>=2.1.0",
     "cryptography>=46.0.3",
+    "gradio>=4.0.0",
+    "pandas>=2.0.0",
+    "plotly>=5.0.0",
 ]

scripts/README.md

@@ -0,0 +1,78 @@
+# Migration Scripts
+
+Utility scripts per la gestione della migrazione.
+
+## sync_migration_state.py
+
+Sincronizza la tabella `migration_state` con i dati effettivamente presenti in PostgreSQL.
+
+### Quando usare
+
+Usa questo script quando `migration_state` non è sincronizzato con i dati reali, ad esempio:
+- Dopo inserimenti manuali in PostgreSQL
+- Dopo corruzione dello stato
+- Prima di eseguire migrazione incrementale su dati già esistenti
+
+### Come funziona
+
+Per ogni tabella (rawdatacor, elabdatadisp):
+1. Trova la riga con MAX(created_at) - l'ultima riga inserita
+2. Estrae la consolidation key da quella riga
+3. Aggiorna `migration_state._global` con quella chiave
+
+### Utilizzo
+
+```bash
+# Eseguire dalla root del progetto
+python scripts/sync_migration_state.py
+```
+
+### Output
+
+```
+Syncing migration_state with actual PostgreSQL data...
+================================================================================
+
+ELABDATADISP:
+  Most recently inserted row (by created_at):
+    created_at: 2025-12-30 11:58:24
+    event_timestamp: 2025-12-30 14:58:24
+    Consolidation key: (ID0290, DT0007, 2025-12-30, 14:58:24)
+  ✓ Updated migration_state with this key
+
+RAWDATACOR:
+  Most recently inserted row (by created_at):
+    created_at: 2025-12-30 11:13:29
+    event_timestamp: 2025-12-30 11:11:39
+    Consolidation key: (ID0304, DT0024, 2025-12-30, 11:11:39)
+  ✓ Updated migration_state with this key
+
+================================================================================
+✓ Done! Incremental migration will now start from the correct position.
+```
+
+### Effetti
+
+Dopo aver eseguito questo script:
+- `migration_state._global` sarà aggiornato con l'ultima chiave migrata
+- `python main.py migrate incremental` partirà dalla posizione corretta
+- Non verranno create duplicazioni (usa ON CONFLICT DO NOTHING)
+
+### Avvertenze
+
+- Esclude automaticamente dati corrotti (unit_name come `[Ljava.lang.String;@...`)
+- Usa `created_at` per trovare l'ultima riga inserita (non `event_timestamp`)
+- Sovrascrive lo stato globale esistente
+
+### Verifica
+
+Dopo aver eseguito lo script, verifica lo stato:
+
+```sql
+SELECT table_name, partition_name, last_key
+FROM migration_state
+WHERE partition_name = '_global'
+ORDER BY table_name;
+```
+
+Dovrebbe mostrare le chiavi più recenti per entrambe le tabelle.

scripts/sync_migration_state.py

@@ -0,0 +1,63 @@
+#!/usr/bin/env python3
+"""Sync migration_state with actual data in PostgreSQL tables."""
+
+import sys
+sys.path.insert(0, '/home/alex/devel/mysql2postgres')
+
+from src.connectors.postgres_connector import PostgreSQLConnector
+from src.migrator.state_manager import StateManager
+
+def sync_table_state(table_name: str):
+    """Sync migration_state for a table with its actual data."""
+    with PostgreSQLConnector() as pg_conn:
+        cursor = pg_conn.connection.cursor()
+
+        # Find the row with MAX(created_at) - most recently inserted
+        # Exclude corrupted data (Java strings)
+        cursor.execute(f"""
+            SELECT unit_name, tool_name_id,
+                   DATE(event_timestamp)::text as event_date,
+                   event_timestamp::time::text as event_time,
+                   created_at,
+                   event_timestamp
+            FROM {table_name}
+            WHERE unit_name NOT LIKE '[L%'  -- Exclude corrupted Java strings
+            ORDER BY created_at DESC
+            LIMIT 1
+        """)
+
+        result = cursor.fetchone()
+        if not result:
+            print(f"No data found in {table_name}")
+            return
+
+        unit_name, tool_name_id, event_date, event_time, created_at, event_timestamp = result
+
+        print(f"\n{table_name.upper()}:")
+        print(f"  Most recently inserted row (by created_at):")
+        print(f"    created_at: {created_at}")
+        print(f"    event_timestamp: {event_timestamp}")
+        print(f"    Consolidation key: ({unit_name}, {tool_name_id}, {event_date}, {event_time})")
+
+        # Update global migration_state with this key
+        state_mgr = StateManager(pg_conn, table_name, partition_name="_global")
+
+        last_key = {
+            "unit_name": unit_name,
+            "tool_name_id": tool_name_id,
+            "event_date": event_date,
+            "event_time": event_time
+        }
+
+        state_mgr.update_state(last_key=last_key)
+        print(f"  ✓ Updated migration_state with this key")
+
+if __name__ == "__main__":
+    print("Syncing migration_state with actual PostgreSQL data...")
+    print("="*80)
+
+    sync_table_state("elabdatadisp")
+    sync_table_state("rawdatacor")
+
+    print("\n" + "="*80)
+    print("✓ Done! Incremental migration will now start from the correct position.")

src/connectors/mysql_connector.py

@@ -1,6 +1,6 @@
 """MySQL database connector."""
 import pymysql
-from typing import List, Dict, Any, Optional, Generator
+from typing import List, Dict, Any, Optional, Generator, Iterator
 from config import get_settings
 from src.utils.logger import get_logger
 
@@ -10,6 +10,8 @@ logger = get_logger(__name__)
 class MySQLConnector:
     """Connector for MySQL database."""
 
+    MAX_RETRIES = 3  # Number of retries for transient connection errors
+
     def __init__(self):
         """Initialize MySQL connector with settings."""
         self.settings = get_settings()
@@ -38,6 +40,16 @@ class MySQLConnector:
             logger.error(f"Failed to connect to MySQL: {e}")
             raise
 
+    def _reconnect(self) -> None:
+        """Reconnect to MySQL database after connection loss."""
+        try:
+            self.disconnect()
+            self.connect()
+            logger.info("Successfully reconnected to MySQL")
+        except Exception as e:
+            logger.error(f"Failed to reconnect to MySQL: {e}")
+            raise
+
     def disconnect(self) -> None:
         """Close connection to MySQL database."""
         if self.connection:
@@ -62,14 +74,21 @@ class MySQLConnector:
         Returns:
             Number of rows in the table
         """
-        try:
-            with self.connection.cursor() as cursor:
-                cursor.execute(f"SELECT COUNT(*) as count FROM `{table}`")
-                result = cursor.fetchone()
-                return result["count"]
-        except pymysql.Error as e:
-            logger.error(f"Failed to get row count for {table}: {e}")
-            raise
+        retries = 0
+        while retries < self.MAX_RETRIES:
+            try:
+                with self.connection.cursor() as cursor:
+                    cursor.execute(f"SELECT COUNT(*) as count FROM `{table}`")
+                    result = cursor.fetchone()
+                    return result["count"]
+            except pymysql.Error as e:
+                retries += 1
+                if retries >= self.MAX_RETRIES:
+                    logger.error(f"Failed to get row count for {table} after {self.MAX_RETRIES} retries: {e}")
+                    raise
+                else:
+                    logger.warning(f"Get row count failed (retry {retries}/{self.MAX_RETRIES}): {e}")
+                    self._reconnect()
 
     def fetch_rows_since(
         self,
@@ -195,19 +214,26 @@ class MySQLConnector:
         Returns:
             List of partition names
         """
-        try:
-            with self.connection.cursor() as cursor:
-                cursor.execute("""
-                    SELECT PARTITION_NAME
-                    FROM INFORMATION_SCHEMA.PARTITIONS
-                    WHERE TABLE_NAME = %s
-                      AND TABLE_SCHEMA = %s
-                    ORDER BY PARTITION_NAME
-                """, (table, self.settings.mysql.database))
-                return [row["PARTITION_NAME"] for row in cursor.fetchall()]
-        except pymysql.Error as e:
-            logger.error(f"Failed to get partitions for {table}: {e}")
-            raise
+        retries = 0
+        while retries < self.MAX_RETRIES:
+            try:
+                with self.connection.cursor() as cursor:
+                    cursor.execute("""
+                        SELECT PARTITION_NAME
+                        FROM INFORMATION_SCHEMA.PARTITIONS
+                        WHERE TABLE_NAME = %s
+                          AND TABLE_SCHEMA = %s
+                        ORDER BY PARTITION_NAME
+                    """, (table, self.settings.mysql.database))
+                    return [row["PARTITION_NAME"] for row in cursor.fetchall()]
+            except pymysql.Error as e:
+                retries += 1
+                if retries >= self.MAX_RETRIES:
+                    logger.error(f"Failed to get partitions for {table} after {self.MAX_RETRIES} retries: {e}")
+                    raise
+                else:
+                    logger.warning(f"Get table partitions failed (retry {retries}/{self.MAX_RETRIES}): {e}")
+                    self._reconnect()
 
     def fetch_consolidation_groups_from_partition(
         self,
@@ -215,7 +241,8 @@ class MySQLConnector:
         partition: str,
         limit: Optional[int] = None,
         offset: int = 0,
-        start_id: Optional[int] = None
+        start_id: Optional[int] = None,
+        start_key: Optional[tuple] = None
     ) -> Generator[List[Dict[str, Any]], None, None]:
         """Fetch consolidation groups from a partition.
 
@@ -231,6 +258,8 @@ class MySQLConnector:
             limit: Batch size for consolidation (uses config default if None)
             offset: Starting offset for pagination (unused, kept for compatibility)
             start_id: Resume from this ID (fetch id > start_id). If None, starts from beginning
+            start_key: Resume AFTER this consolidation key (unit_name, tool_name_id, event_date, event_time).
+                      If provided, skips all keys <= start_key in MySQL query (efficient resume).
 
         Yields:
             Lists of rows grouped by consolidation key (complete groups only)
@@ -243,47 +272,53 @@ class MySQLConnector:
 
         # Determine ID column name
         id_column = "idElabData" if table == "ELABDATADISP" else "id"
-        max_retries = 3
-        last_key = None
+
+        # Initialize last_completed_key with start_key for efficient resume
+        last_completed_key = start_key if start_key else None
+
+        # CRITICAL: These must be OUTSIDE the while loop to persist across batch iterations
+        # If they're inside the loop, buffered incomplete groups from previous batches get lost
+        current_group = []
+        current_key = None
 
         while True:
             retries = 0
-            while retries < max_retries:
+            while retries < self.MAX_RETRIES:
                 try:
                     with self.connection.cursor() as cursor:
-                        # ORDER BY consolidation key (without NodeNum for speed)
-                        # Ensures all rows of a key arrive together, then we sort by NodeNum in memory
-                        if last_key is None:
+                        # ORDER BY consolidation key
+                        # Ensures all rows of a key arrive together
+                        if last_completed_key is None:
                             rows_query = f"""
                                 SELECT * FROM `{table}` PARTITION (`{partition}`)
                                 ORDER BY UnitName ASC, ToolNameID ASC, EventDate ASC, EventTime ASC
                                 LIMIT %s
                             """
+                            logger.info(f"Executing first query on partition {partition} (fetching up to {limit} rows, sorted by consolidation key)...")
                             cursor.execute(rows_query, (limit,))
                         else:
-                            # Resume after last key using tuple comparison
+                            # Resume AFTER last completely yielded key
+                            # This ensures we don't re-process any rows
                             rows_query = f"""
                                 SELECT * FROM `{table}` PARTITION (`{partition}`)
                                 WHERE (UnitName, ToolNameID, EventDate, EventTime) > (%s, %s, %s, %s)
                                 ORDER BY UnitName ASC, ToolNameID ASC, EventDate ASC, EventTime ASC
                                 LIMIT %s
                             """
-                            cursor.execute(rows_query, (last_key[0], last_key[1], last_key[2], last_key[3], limit))
+                            logger.debug(f"Executing query on partition {partition} (resuming from key {last_completed_key}, limit {limit})...")
+                            cursor.execute(rows_query, (last_completed_key[0], last_completed_key[1], last_completed_key[2], last_completed_key[3], limit))
 
                         rows = cursor.fetchall()
+                        logger.debug(f"Fetched {len(rows)} rows from partition {partition}")
 
                         if not rows:
+                            # No more rows - yield any buffered group and finish
+                            if current_group:
+                                yield current_group
                             return
 
-                        # DO NOT re-sort by NodeNum! The database has already ordered by consolidation key,
-                        # and re-sorting breaks the grouping. Rows with same key will not be consecutive anymore.
-                        # We trust the MySQL ORDER BY to keep all rows of same key together.
-
-                        # Group rows by consolidation key
-                        # Since rows are already ordered by key from MySQL ORDER BY, all rows with same key are consecutive
-                        current_group = []
-                        current_key = None
-
+                        # Process batch: add rows to groups, yield only COMPLETE groups
+                        # Keep incomplete groups buffered in current_group for next batch
                         for row in rows:
                             key = (
                                 row.get("UnitName"),
@@ -292,44 +327,513 @@ class MySQLConnector:
                                 row.get("EventTime")
                             )
 
-                            # Yield group when key changes
+                            # Key changed - previous group is complete, yield it
                             if current_key is not None and key != current_key:
-                                if current_group:
-                                    yield current_group
+                                yield current_group
                                 current_group = []
+                                last_completed_key = current_key
 
                             current_group.append(row)
                             current_key = key
 
-                        # At end of batch: handle final group
-                        if not rows or len(rows) < limit:
+                        # At batch boundary: only yield if group is DEFINITELY complete
+                        # A group is incomplete if the last row has same key as current group
+                        if rows and len(rows) < limit:
                             # Last batch - yield remaining group and finish
                             if current_group:
                                 yield current_group
                             return
-                        else:
-                            # More rows might exist - yield the last group only if key changed
-                            # If not, it will be continued/merged in next iteration
-                            if current_group:
-                                yield current_group
+                        elif rows:
+                            # More rows exist - check if last row belongs to current group
+                            last_row_key = (
+                                rows[-1].get("UnitName"),
+                                rows[-1].get("ToolNameID"),
+                                rows[-1].get("EventDate"),
+                                rows[-1].get("EventTime")
+                            )
 
-                        # Update last_key for next iteration
-                        if current_key:
-                            last_key = current_key
+                            if last_row_key != current_key:
+                                # Last row was from a previous (complete) group - shouldn't happen
+                                # given the ORDER BY, but if it does, we've already yielded it above
+                                pass
+                            # else: current group is incomplete (extends past this batch)
+                            # Keep it buffered for next batch
 
                         break  # Success, exit retry loop
 
                 except pymysql.Error as e:
                     retries += 1
-                    if retries >= max_retries:
-                        logger.error(f"Failed to fetch consolidation groups from {table} partition {partition} after {max_retries} retries: {e}")
+                    if retries >= self.MAX_RETRIES:
+                        logger.error(f"Failed to fetch consolidation groups from {table} partition {partition} after {self.MAX_RETRIES} retries: {e}")
                         raise
                     else:
-                        logger.warning(f"Fetch failed (retry {retries}/{max_retries}): {e}")
-                        # Reconnect and retry
-                        try:
-                            self.disconnect()
-                            self.connect()
-                        except Exception as reconnect_error:
-                            logger.error(f"Failed to reconnect: {reconnect_error}")
-                            raise
+                        logger.warning(f"Fetch consolidation groups failed (retry {retries}/{self.MAX_RETRIES}): {e}")
+                        self._reconnect()
+
+    def stream_partition_rows(
+        self,
+        table: str,
+        partition: str,
+        batch_size: int = 10000,
+        resume_from_key: Optional[Dict[str, Any]] = None
+    ) -> Iterator[Dict[str, Any]]:
+        """Stream all rows from a partition ordered by consolidation key.
+
+        This eliminates the N+1 query pattern by fetching all data in a single
+        streaming query ordered by consolidation key. Rows are yielded one at a time.
+
+        Query pattern:
+            SELECT *
+            FROM table PARTITION (partition)
+            WHERE (UnitName, ToolNameID, EventDate, EventTime) > (?, ?, ?, ?)  -- if resuming
+            ORDER BY UnitName, ToolNameID, EventDate, EventTime, NodeNum
+
+        Args:
+            table: Table name (RAWDATACOR or ELABDATADISP)
+            partition: Partition name (e.g., 'p2024')
+            batch_size: Number of rows to fetch at a time from server
+            resume_from_key: Last processed key to resume from (optional)
+                Format: {"unit_name": ..., "tool_name_id": ..., "event_date": ..., "event_time": ...}
+
+        Yields:
+            MySQL row dicts, ordered by consolidation key then NodeNum
+        """
+        if table not in ("RAWDATACOR", "ELABDATADISP"):
+            raise ValueError(f"Consolidation not supported for table {table}")
+
+        retries = 0
+        while retries < self.MAX_RETRIES:
+            try:
+                # Use regular DictCursor (client-side) to avoid filling /tmp on MySQL server
+                # SSCursor creates temp files on MySQL server which can fill /tmp
+                with self.connection.cursor() as cursor:
+                    # Build WHERE clause for resume
+                    where_clause = ""
+                    params = []
+
+                    if resume_from_key:
+                        # Resume from last key using tuple comparison
+                        where_clause = """
+                            WHERE (UnitName, ToolNameID, EventDate, EventTime) > (%s, %s, %s, %s)
+                        """
+                        params = [
+                            resume_from_key.get("unit_name"),
+                            resume_from_key.get("tool_name_id"),
+                            resume_from_key.get("event_date"),
+                            resume_from_key.get("event_time"),
+                        ]
+
+                    query = f"""
+                        SELECT *
+                        FROM `{table}` PARTITION (`{partition}`)
+                        {where_clause}
+                        ORDER BY UnitName, ToolNameID, EventDate, EventTime, NodeNum
+                    """
+
+                    logger.info(
+                        f"Starting stream from {table} partition {partition}"
+                        + (f" (resuming from key)" if resume_from_key else "")
+                    )
+
+                    cursor.execute(query, tuple(params) if params else None)
+
+                    # Fetch all rows (client-side cursor fetches to local memory)
+                    # This avoids creating temp files on MySQL server
+                    rows = cursor.fetchall()
+
+                    logger.info(f"Fetched {len(rows)} rows from partition {partition}")
+
+                    # Yield rows in batches for memory efficiency
+                    for i in range(0, len(rows), batch_size):
+                        batch = rows[i:i+batch_size]
+                        for row in batch:
+                            yield row
+
+                    logger.info(f"Finished streaming partition {partition}")
+                    return  # Success
+
+            except pymysql.Error as e:
+                retries += 1
+                if retries >= self.MAX_RETRIES:
+                    logger.error(
+                        f"Failed to stream from {table} partition {partition} "
+                        f"after {self.MAX_RETRIES} retries: {e}"
+                    )
+                    raise
+                else:
+                    logger.warning(
+                        f"Stream failed (retry {retries}/{self.MAX_RETRIES}): {e}"
+                    )
+                    self._reconnect()
+
+    def fetch_consolidation_keys_from_partition(
+        self,
+        table: str,
+        partition: str,
+        limit: Optional[int] = None,
+        offset: int = 0
+    ) -> List[Dict[str, Any]]:
+        """Fetch distinct consolidation keys from a partition.
+
+        Query pattern:
+            SELECT UnitName, ToolNameID, EventDate, EventTime
+            FROM table PARTITION (partition)
+            GROUP BY UnitName, ToolNameID, EventDate, EventTime
+            ORDER BY UnitName, ToolNameID, EventDate, EventTime
+            LIMIT X OFFSET Y
+
+        Args:
+            table: Table name (RAWDATACOR or ELABDATADISP)
+            partition: Partition name (e.g., 'p2024')
+            limit: Number of keys to fetch (uses CONSOLIDATION_GROUP_LIMIT if None)
+            offset: Starting offset for pagination
+
+        Returns:
+            List of dicts with keys: UnitName, ToolNameID, EventDate, EventTime
+        """
+        if table not in ("RAWDATACOR", "ELABDATADISP"):
+            raise ValueError(f"Consolidation not supported for table {table}")
+
+        if limit is None:
+            limit = self.settings.migration.consolidation_group_limit
+
+        retries = 0
+        while retries < self.MAX_RETRIES:
+            try:
+                with self.connection.cursor() as cursor:
+                    query = f"""
+                        SELECT UnitName, ToolNameID, EventDate, EventTime
+                        FROM `{table}` PARTITION (`{partition}`)
+                        GROUP BY UnitName, ToolNameID, EventDate, EventTime
+                        ORDER BY UnitName, ToolNameID, EventDate, EventTime
+                        LIMIT %s OFFSET %s
+                    """
+                    cursor.execute(query, (limit, offset))
+                    keys = cursor.fetchall()
+                    return keys
+
+            except pymysql.Error as e:
+                retries += 1
+                if retries >= self.MAX_RETRIES:
+                    logger.error(
+                        f"Failed to fetch consolidation keys from {table} "
+                        f"partition {partition} (offset={offset}) after {self.MAX_RETRIES} retries: {e}"
+                    )
+                    raise
+                else:
+                    logger.warning(
+                        f"Fetch consolidation keys failed (retry {retries}/{self.MAX_RETRIES}): {e}"
+                    )
+                    self._reconnect()
+
+    def fetch_records_for_key(
+        self,
+        table: str,
+        partition: str,
+        unit_name: Any,
+        tool_name_id: Any,
+        event_date: Any,
+        event_time: Any
+    ) -> List[Dict[str, Any]]:
+        """Fetch all records for a specific consolidation key.
+
+        Query pattern:
+            SELECT *
+            FROM table PARTITION (partition)
+            WHERE UnitName = ? AND ToolNameID = ?
+              AND EventDate = ? AND EventTime = ?
+
+        Args:
+            table: Table name (RAWDATACOR or ELABDATADISP)
+            partition: Partition name
+            unit_name: UnitName value
+            tool_name_id: ToolNameID value
+            event_date: EventDate value
+            event_time: EventTime value
+
+        Returns:
+            List of all MySQL rows matching the key (different NodeNum)
+        """
+        if table not in ("RAWDATACOR", "ELABDATADISP"):
+            raise ValueError(f"Consolidation not supported for table {table}")
+
+        retries = 0
+        while retries < self.MAX_RETRIES:
+            try:
+                with self.connection.cursor() as cursor:
+                    query = f"""
+                        SELECT *
+                        FROM `{table}` PARTITION (`{partition}`)
+                        WHERE UnitName = %s
+                          AND ToolNameID = %s
+                          AND EventDate = %s
+                          AND EventTime = %s
+                        ORDER BY NodeNum
+                    """
+                    cursor.execute(query, (unit_name, tool_name_id, event_date, event_time))
+                    rows = cursor.fetchall()
+                    return rows
+
+            except pymysql.Error as e:
+                retries += 1
+                if retries >= self.MAX_RETRIES:
+                    logger.error(
+                        f"Failed to fetch records for key "
+                        f"({unit_name}, {tool_name_id}, {event_date}, {event_time}) "
+                        f"from {table} partition {partition} after {self.MAX_RETRIES} retries: {e}"
+                    )
+                    raise
+                else:
+                    logger.warning(
+                        f"Fetch records for key failed (retry {retries}/{self.MAX_RETRIES}): {e}"
+                    )
+                    self._reconnect()
+
+    def fetch_consolidation_keys_after(
+        self,
+        table: str,
+        after_key: Optional[Dict[str, Any]] = None,
+        min_mysql_id: int = 0,
+        limit: Optional[int] = None,
+        offset: int = 0
+    ) -> List[Dict[str, Any]]:
+        """Fetch distinct consolidation keys after a specific key (for incremental migration).
+
+        Query pattern:
+            SELECT UnitName, ToolNameID, EventDate, EventTime
+            FROM table
+            WHERE id > min_mysql_id
+              AND (UnitName, ToolNameID, EventDate, EventTime) > (?, ?, ?, ?)
+            GROUP BY UnitName, ToolNameID, EventDate, EventTime
+            ORDER BY UnitName, ToolNameID, EventDate, EventTime
+            LIMIT X OFFSET Y
+
+        Args:
+            table: Table name (RAWDATACOR or ELABDATADISP)
+            after_key: Start after this key (dict with unit_name, tool_name_id, event_date, event_time)
+            min_mysql_id: Only fetch rows with id > this value (optimization to avoid scanning already migrated data)
+            limit: Number of keys to fetch (uses CONSOLIDATION_GROUP_LIMIT if None)
+            offset: Starting offset for pagination
+
+        Returns:
+            List of dicts with keys: UnitName, ToolNameID, EventDate, EventTime
+        """
+        if table not in ("RAWDATACOR", "ELABDATADISP"):
+            raise ValueError(f"Consolidation not supported for table {table}")
+
+        if limit is None:
+            limit = self.settings.migration.consolidation_group_limit
+
+        # Determine ID column name based on table
+        id_column = "idElabData" if table == "ELABDATADISP" else "id"
+
+        retries = 0
+        while retries < self.MAX_RETRIES:
+            try:
+                with self.connection.cursor() as cursor:
+                    if after_key:
+                        # Incremental: fetch keys AFTER the last migrated key
+                        # Filter by ID first for performance (uses PRIMARY KEY index)
+                        query = f"""
+                            SELECT UnitName, ToolNameID, EventDate, EventTime
+                            FROM `{table}`
+                            WHERE `{id_column}` > %s
+                              AND (UnitName, ToolNameID, EventDate, EventTime) > (%s, %s, %s, %s)
+                            GROUP BY UnitName, ToolNameID, EventDate, EventTime
+                            ORDER BY UnitName, ToolNameID, EventDate, EventTime
+                            LIMIT %s OFFSET %s
+                        """
+                        cursor.execute(
+                            query,
+                            (
+                                min_mysql_id,
+                                after_key.get("unit_name"),
+                                after_key.get("tool_name_id"),
+                                after_key.get("event_date"),
+                                after_key.get("event_time"),
+                                limit,
+                                offset
+                            )
+                        )
+                    else:
+                        # No after_key: fetch from beginning
+                        # Still filter by ID if min_mysql_id is provided
+                        if min_mysql_id > 0:
+                            query = f"""
+                                SELECT UnitName, ToolNameID, EventDate, EventTime
+                                FROM `{table}`
+                                WHERE `{id_column}` > %s
+                                GROUP BY UnitName, ToolNameID, EventDate, EventTime
+                                ORDER BY UnitName, ToolNameID, EventDate, EventTime
+                                LIMIT %s OFFSET %s
+                            """
+                            cursor.execute(query, (min_mysql_id, limit, offset))
+                        else:
+                            query = f"""
+                                SELECT UnitName, ToolNameID, EventDate, EventTime
+                                FROM `{table}`
+                                GROUP BY UnitName, ToolNameID, EventDate, EventTime
+                                ORDER BY UnitName, ToolNameID, EventDate, EventTime
+                                LIMIT %s OFFSET %s
+                            """
+                            cursor.execute(query, (limit, offset))
+
+                    keys = cursor.fetchall()
+                    return keys
+
+            except pymysql.Error as e:
+                retries += 1
+                if retries >= self.MAX_RETRIES:
+                    logger.error(
+                        f"Failed to fetch consolidation keys from {table} "
+                        f"(after_key={after_key}, offset={offset}) after {self.MAX_RETRIES} retries: {e}"
+                    )
+                    raise
+                else:
+                    logger.warning(
+                        f"Fetch consolidation keys after failed (retry {retries}/{self.MAX_RETRIES}): {e}"
+                    )
+                    self._reconnect()
+
+    def fetch_consolidation_keys_from_partition_after(
+        self,
+        table: str,
+        partition: str,
+        after_key: Optional[Dict[str, Any]] = None,
+        limit: Optional[int] = None
+    ) -> List[Dict[str, Any]]:
+        """Fetch distinct consolidation keys from a specific partition after a specific key.
+
+        Optimized version for incremental migration that queries only one partition.
+
+        Query pattern:
+            SELECT UnitName, ToolNameID, EventDate, EventTime
+            FROM table PARTITION (partition_name)
+            WHERE (UnitName, ToolNameID, EventDate, EventTime) > (?, ?, ?, ?)
+            GROUP BY UnitName, ToolNameID, EventDate, EventTime
+            ORDER BY UnitName, ToolNameID, EventDate, EventTime
+            LIMIT X
+
+        Args:
+            table: Table name (RAWDATACOR or ELABDATADISP)
+            partition: Partition name (e.g., 'part8', 'd9')
+            after_key: Start after this key (dict with unit_name, tool_name_id, event_date, event_time)
+            limit: Number of keys to fetch (uses CONSOLIDATION_GROUP_LIMIT if None)
+
+        Returns:
+            List of dicts with keys: UnitName, ToolNameID, EventDate, EventTime
+        """
+        if table not in ("RAWDATACOR", "ELABDATADISP"):
+            raise ValueError(f"Consolidation not supported for table {table}")
+
+        if limit is None:
+            limit = self.settings.migration.consolidation_group_limit
+
+        retries = 0
+        while retries < self.MAX_RETRIES:
+            try:
+                with self.connection.cursor() as cursor:
+                    if after_key:
+                        # Fetch keys AFTER the last migrated key from this specific partition
+                        query = f"""
+                            SELECT UnitName, ToolNameID, EventDate, EventTime
+                            FROM `{table}` PARTITION (`{partition}`)
+                            WHERE (UnitName, ToolNameID, EventDate, EventTime) > (%s, %s, %s, %s)
+                            GROUP BY UnitName, ToolNameID, EventDate, EventTime
+                            ORDER BY UnitName, ToolNameID, EventDate, EventTime
+                            LIMIT %s
+                        """
+                        cursor.execute(
+                            query,
+                            (
+                                after_key.get("unit_name"),
+                                after_key.get("tool_name_id"),
+                                after_key.get("event_date"),
+                                after_key.get("event_time"),
+                                limit
+                            )
+                        )
+                    else:
+                        # No after_key: fetch from beginning of partition
+                        query = f"""
+                            SELECT UnitName, ToolNameID, EventDate, EventTime
+                            FROM `{table}` PARTITION (`{partition}`)
+                            GROUP BY UnitName, ToolNameID, EventDate, EventTime
+                            ORDER BY UnitName, ToolNameID, EventDate, EventTime
+                            LIMIT %s
+                        """
+                        cursor.execute(query, (limit,))
+
+                    keys = cursor.fetchall()
+                    return keys
+
+            except pymysql.Error as e:
+                retries += 1
+                if retries >= self.MAX_RETRIES:
+                    logger.error(
+                        f"Failed to fetch consolidation keys from {table} PARTITION ({partition}) "
+                        f"(after_key={after_key}) after {self.MAX_RETRIES} retries: {e}"
+                    )
+                    raise
+                else:
+                    logger.warning(
+                        f"Fetch consolidation keys from partition failed (retry {retries}/{self.MAX_RETRIES}): {e}"
+                    )
+                    self._reconnect()
+
+    def fetch_records_for_key_all_partitions(
+        self,
+        table: str,
+        unit_name: Any,
+        tool_name_id: Any,
+        event_date: Any,
+        event_time: Any
+    ) -> List[Dict[str, Any]]:
+        """Fetch all records for a specific consolidation key across all partitions.
+
+        Used for incremental migration where we don't know which partition the key is in.
+
+        Args:
+            table: Table name (RAWDATACOR or ELABDATADISP)
+            unit_name: UnitName value
+            tool_name_id: ToolNameID value
+            event_date: EventDate value
+            event_time: EventTime value
+
+        Returns:
+            List of all MySQL rows matching the key (different NodeNum)
+        """
+        if table not in ("RAWDATACOR", "ELABDATADISP"):
+            raise ValueError(f"Consolidation not supported for table {table}")
+
+        retries = 0
+        while retries < self.MAX_RETRIES:
+            try:
+                with self.connection.cursor() as cursor:
+                    query = f"""
+                        SELECT *
+                        FROM `{table}`
+                        WHERE UnitName = %s
+                          AND ToolNameID = %s
+                          AND EventDate = %s
+                          AND EventTime = %s
+                        ORDER BY NodeNum
+                    """
+                    cursor.execute(query, (unit_name, tool_name_id, event_date, event_time))
+                    rows = cursor.fetchall()
+                    return rows
+
+            except pymysql.Error as e:
+                retries += 1
+                if retries >= self.MAX_RETRIES:
+                    logger.error(
+                        f"Failed to fetch records for key "
+                        f"({unit_name}, {tool_name_id}, {event_date}, {event_time}) "
+                        f"from {table} after {self.MAX_RETRIES} retries: {e}"
+                    )
+                    raise
+                else:
+                    logger.warning(
+                        f"Fetch records for key (all partitions) failed (retry {retries}/{self.MAX_RETRIES}): {e}"
+                    )
+                    self._reconnect()

src/connectors/postgres_connector.py

@@ -1,7 +1,6 @@
 """PostgreSQL database connector."""
 import psycopg
-from typing import List, Dict, Any, Optional, Iterator
-from psycopg import sql
+from typing import List, Dict, Any
 import json
 from config import get_settings
 from src.utils.logger import get_logger
@@ -13,72 +12,151 @@ class PostgreSQLConnector:
     """Connector for PostgreSQL database."""
 
     def __init__(self):
-        """Initialize PostgreSQL connector with settings."""
+        """Initialize the PostgreSQL connector."""
         self.settings = get_settings()
         self.connection = None
 
-    def connect(self) -> None:
-        """Establish connection to PostgreSQL database."""
+    def connect(self):
+        """Connect to PostgreSQL."""
         try:
             self.connection = psycopg.connect(
                 host=self.settings.postgres.host,
                 port=self.settings.postgres.port,
+                dbname=self.settings.postgres.database,
                 user=self.settings.postgres.user,
                 password=self.settings.postgres.password,
-                dbname=self.settings.postgres.database,
                 autocommit=False,
             )
             logger.info(
-                f"Connected to PostgreSQL: {self.settings.postgres.host}:"
-                f"{self.settings.postgres.port}/{self.settings.postgres.database}"
+                f"Connected to PostgreSQL: {self.settings.postgres.host}:{self.settings.postgres.port}/{self.settings.postgres.database}"
             )
         except psycopg.Error as e:
             logger.error(f"Failed to connect to PostgreSQL: {e}")
             raise
 
-    def disconnect(self) -> None:
-        """Close connection to PostgreSQL database."""
+    def disconnect(self):
+        """Disconnect from PostgreSQL."""
         if self.connection:
             self.connection.close()
             logger.info("Disconnected from PostgreSQL")
 
     def __enter__(self):
-        """Context manager entry."""
+        """Context manager entry - connect to database."""
         self.connect()
         return self
 
     def __exit__(self, exc_type, exc_val, exc_tb):
-        """Context manager exit."""
-        if exc_type is None:
-            # No exception, commit before closing
-            try:
-                self.connection.commit()
-            except Exception as e:
-                logger.warning(f"Failed to commit on exit: {e}")
-        else:
-            # Exception occurred, rollback
-            try:
-                self.connection.rollback()
-            except Exception as e:
-                logger.warning(f"Failed to rollback on exit: {e}")
+        """Context manager exit - disconnect from database."""
         self.disconnect()
+        return False
 
-    def execute(self, query: str, params: Optional[tuple] = None) -> None:
-        """Execute a query without returning results with retry logic.
+    def execute_query(
+        self, query: str, params: tuple = None, fetch: bool = False
+    ) -> List[Dict[str, Any]]:
+        """Execute a query and optionally fetch results.
 
         Args:
-            query: SQL query
+            query: SQL query to execute
             params: Query parameters
+            fetch: Whether to fetch and return results
+
+        Returns:
+            List of result rows as dictionaries (if fetch=True)
         """
+        try:
+            with self.connection.cursor() as cursor:
+                cursor.execute(query, params)
+                if fetch:
+                    columns = [desc[0] for desc in cursor.description]
+                    rows = cursor.fetchall()
+                    return [dict(zip(columns, row)) for row in rows]
+                self.connection.commit()
+                return []
+        except psycopg.Error as e:
+            self.connection.rollback()
+            logger.error(f"Query execution failed: {e}")
+            raise
+
+    def execute_script(self, script: str):
+        """Execute a SQL script (multiple statements).
+
+        Args:
+            script: SQL script to execute
+        """
+        try:
+            with self.connection.cursor() as cursor:
+                cursor.execute(script)
+                self.connection.commit()
+            logger.debug("Script executed successfully")
+        except psycopg.Error as e:
+            self.connection.rollback()
+            logger.error(f"Script execution failed: {e}")
+            raise
+
+    def copy_from(
+        self,
+        table: str,
+        rows: List[Dict[str, Any]],
+        columns: List[str]
+    ) -> int:
+        """Insert a batch of rows using PostgreSQL COPY (10-100x faster than INSERT).
+
+        Args:
+            table: Table name
+            rows: List of row dictionaries
+            columns: Column names in order
+
+        Returns:
+            Number of rows inserted
+        """
+        if not rows:
+            return 0
+
         max_retries = 3
         retries = 0
 
         while retries < max_retries:
             try:
                 with self.connection.cursor() as cursor:
-                    cursor.execute(query, params)
+                    # Prepare data for COPY
+                    # COPY expects tab-separated text with one row per line
+                    from io import StringIO
+
+                    copy_data = StringIO()
+                    for row in rows:
+                        values = []
+                        for col in columns:
+                            val = row.get(col)
+                            # Handle None/NULL
+                            if val is None:
+                                values.append("\\N")
+                            # Convert JSONB dicts to JSON strings
+                            elif isinstance(val, (dict, list)):
+                                # Escape special characters in JSON
+                                json_str = json.dumps(val).replace("\\", "\\\\").replace("\n", "\\n").replace("\r", "\\r").replace("\t", "\\t")
+                                values.append(json_str)
+                            # Convert bool to PostgreSQL format
+                            elif isinstance(val, bool):
+                                values.append("t" if val else "f")
+                            # All other values as strings
+                            else:
+                                # Escape special characters
+                                str_val = str(val).replace("\\", "\\\\").replace("\n", "\\n").replace("\r", "\\r").replace("\t", "\\t")
+                                values.append(str_val)
+
+                        copy_data.write("\t".join(values) + "\n")
+
+                    copy_data.seek(0)
+
+                    # Use COPY command
+                    with cursor.copy(f"COPY {table} ({','.join(columns)}) FROM STDIN") as copy:
+                        copy.write(copy_data.getvalue())
+
                     self.connection.commit()
-                return  # Success
+
+                    logger.debug(f"COPY inserted {len(rows)} rows into {table}")
+                    return len(rows)
+
             except psycopg.Error as e:
                 try:
                     self.connection.rollback()
@@ -87,11 +165,129 @@ class PostgreSQLConnector:
 
                 retries += 1
                 if retries >= max_retries:
-                    logger.error(f"Query execution failed after {max_retries} retries: {e}\nQuery: {query}")
+                    logger.error(f"COPY insert failed after {max_retries} retries: {e}")
                     raise
                 else:
                     logger.warning(
-                        f"Query execution failed (retry {retries}/{max_retries}): {e}. "
+                        f"COPY insert failed (retry {retries}/{max_retries}): {e}. "
+                        f"Reconnecting and retrying..."
+                    )
+                    # Reconnect before retry
+                    try:
+                        self.disconnect()
+                    except Exception:
+                        pass
+                    try:
+                        self.connect()
+                    except Exception as reconnect_error:
+                        logger.error(f"Failed to reconnect: {reconnect_error}")
+                        if retries >= max_retries:
+                            raise
+
+    def copy_from_with_conflict(
+        self,
+        table: str,
+        rows: List[Dict[str, Any]],
+        columns: List[str],
+        conflict_columns: List[str]
+    ) -> int:
+        """Insert a batch of rows using COPY + ON CONFLICT for duplicate handling.
+
+        This method is fast (uses COPY) but handles UNIQUE constraint violations:
+        1. COPY data into a temporary table (fast bulk load)
+        2. INSERT from temp table with ON CONFLICT DO NOTHING (skip duplicates)
+
+        Args:
+            table: Target table name
+            rows: List of row dictionaries
+            columns: Column names in order
+            conflict_columns: Columns that form the UNIQUE constraint
+
+        Returns:
+            Number of rows inserted (excludes skipped duplicates)
+        """
+        if not rows:
+            return 0
+
+        max_retries = 3
+        retries = 0
+
+        while retries < max_retries:
+            try:
+                with self.connection.cursor() as cursor:
+                    # Create temporary table with same structure but without id column
+                    # The id column is auto-generated and shouldn't be in the COPY data
+                    temp_table = f"{table}_temp_{id(rows)}"
+                    cursor.execute(f"""
+                        CREATE TEMP TABLE {temp_table}
+                        (LIKE {table} INCLUDING DEFAULTS EXCLUDING IDENTITY)
+                        ON COMMIT DROP
+                    """)
+                    # Drop the id column from temp table since we don't provide it in COPY
+                    cursor.execute(f"ALTER TABLE {temp_table} DROP COLUMN IF EXISTS id")
+
+                    # COPY into temp table
+                    from io import StringIO
+
+                    copy_data = StringIO()
+                    for row in rows:
+                        values = []
+                        for col in columns:
+                            val = row.get(col)
+                            if val is None:
+                                values.append("\\N")
+                            elif isinstance(val, (dict, list)):
+                                json_str = json.dumps(val).replace("\\", "\\\\").replace("\n", "\\n").replace("\r", "\\r").replace("\t", "\\t")
+                                values.append(json_str)
+                            elif isinstance(val, bool):
+                                values.append("t" if val else "f")
+                            else:
+                                str_val = str(val).replace("\\", "\\\\").replace("\n", "\\n").replace("\r", "\\r").replace("\t", "\\t")
+                                values.append(str_val)
+
+                        copy_data.write("\t".join(values) + "\n")
+
+                    copy_data.seek(0)
+
+                    with cursor.copy(f"COPY {temp_table} ({','.join(columns)}) FROM STDIN") as copy:
+                        copy.write(copy_data.getvalue())
+
+                    # INSERT from temp table with ON CONFLICT
+                    conflict_clause = f"({','.join(conflict_columns)})"
+                    insert_sql = f"""
+                        INSERT INTO {table} ({','.join(columns)})
+                        SELECT {','.join(columns)}
+                        FROM {temp_table}
+                        ON CONFLICT {conflict_clause} DO NOTHING
+                    """
+                    cursor.execute(insert_sql)
+                    inserted_count = cursor.rowcount
+
+                    self.connection.commit()
+
+                    if inserted_count < len(rows):
+                        logger.debug(
+                            f"COPY+ON CONFLICT inserted {inserted_count}/{len(rows)} rows into {table} "
+                            f"({len(rows) - inserted_count} duplicates skipped)"
+                        )
+                    else:
+                        logger.debug(f"COPY+ON CONFLICT inserted {inserted_count} rows into {table}")
+
+                    return inserted_count
+
+            except psycopg.Error as e:
+                try:
+                    self.connection.rollback()
+                except Exception:
+                    pass
+
+                retries += 1
+                if retries >= max_retries:
+                    logger.error(f"COPY+ON CONFLICT insert failed after {max_retries} retries: {e}")
+                    raise
+                else:
+                    logger.warning(
+                        f"COPY+ON CONFLICT insert failed (retry {retries}/{max_retries}): {e}. "
                         f"Reconnecting and retrying..."
                     )
                     try:
@@ -105,22 +301,6 @@ class PostgreSQLConnector:
                         if retries >= max_retries:
                             raise
 
-    def execute_script(self, script: str) -> None:
-        """Execute multiple SQL statements (script).
-
-        Args:
-            script: SQL script with multiple statements
-        """
-        try:
-            with self.connection.cursor() as cursor:
-                cursor.execute(script)
-                self.connection.commit()
-            logger.debug("Script executed successfully")
-        except psycopg.Error as e:
-            self.connection.rollback()
-            logger.error(f"Script execution failed: {e}")
-            raise
-
     def insert_batch(
         self,
         table: str,
@@ -204,48 +384,21 @@ class PostgreSQLConnector:
             table: Table name
 
         Returns:
-            True if table exists, False otherwise
+            True if table exists
         """
         try:
             with self.connection.cursor() as cursor:
                 cursor.execute(
-                    "SELECT EXISTS("
-                    "  SELECT 1 FROM information_schema.tables "
-                    "  WHERE table_name = %s"
-                    ")",
-                    (table,)
+                    "SELECT EXISTS (SELECT 1 FROM information_schema.tables WHERE table_name = %s)",
+                    (table,),
                 )
                 return cursor.fetchone()[0]
         except psycopg.Error as e:
-            logger.error(f"Failed to check if table exists: {e}")
-            raise
-
-    def get_max_timestamp(
-        self,
-        table: str,
-        timestamp_col: str = "created_at"
-    ) -> Optional[str]:
-        """Get the maximum timestamp from a table.
-
-        Args:
-            table: Table name
-            timestamp_col: Timestamp column name
-
-        Returns:
-            ISO format timestamp or None if table is empty
-        """
-        try:
-            with self.connection.cursor() as cursor:
-                query = f"SELECT MAX({timestamp_col})::text FROM {table}"
-                cursor.execute(query)
-                result = cursor.fetchone()
-                return result[0] if result and result[0] else None
-        except psycopg.Error as e:
-            logger.error(f"Failed to get max timestamp: {e}")
+            logger.error(f"Failed to check table existence: {e}")
             raise
 
     def get_row_count(self, table: str) -> int:
-        """Get row count for a table.
+        """Get the number of rows in a table.
 
         Args:
             table: Table name

src/migrator/__init__.py

@@ -0,0 +1,25 @@
+"""Migration modules for MySQL to PostgreSQL migration.
+
+New architecture:
+- consolidator: Consolidate MySQL rows by NodeNum into single PostgreSQL row
+- state_manager: Track migration progress in PostgreSQL migration_state table
+- partition_migrator: Migrate single MySQL partition
+- full_migrator: Orchestrate full migration of all partitions
+- incremental_migrator: Migrate only new data since last migration
+"""
+from src.migrator.consolidator import Consolidator, consolidate_rows
+from src.migrator.state_manager import StateManager
+from src.migrator.partition_migrator import PartitionMigrator
+from src.migrator.full_migrator import FullMigrator, run_full_migration
+from src.migrator.incremental_migrator import IncrementalMigrator, run_incremental_migration
+
+__all__ = [
+    "Consolidator",
+    "consolidate_rows",
+    "StateManager",
+    "PartitionMigrator",
+    "FullMigrator",
+    "run_full_migration",
+    "IncrementalMigrator",
+    "run_incremental_migration",
+]

src/migrator/consolidator.py

@@ -0,0 +1,330 @@
+"""Consolidation logic for MySQL to PostgreSQL migration.
+
+Consolidates multiple MySQL rows (different NodeNum) with same consolidation key
+into a single PostgreSQL row with measurements stored in JSONB.
+
+Consolidation key: (UnitName, ToolNameID, EventDate, EventTime)
+Multiple nodes → single JSONB structure with node_X keys
+"""
+from typing import List, Dict, Any, Optional
+from datetime import datetime, date, time, timedelta
+from decimal import Decimal
+
+from config import (
+    TABLE_CONFIGS,
+    RAWDATACOR_COLUMNS,
+    ELABDATADISP_FIELD_MAPPING,
+)
+from src.utils.logger import get_logger
+
+logger = get_logger(__name__)
+
+
+def _combine_datetime(event_date: Any, event_time: Any) -> Optional[datetime]:
+    """Combine MySQL EventDate and EventTime into a single datetime.
+
+    Args:
+        event_date: MySQL EventDate (date object or string)
+        event_time: MySQL EventTime (time object or string)
+
+    Returns:
+        Combined datetime or None if inputs are invalid
+    """
+    if event_date is None or event_time is None:
+        return None
+
+    try:
+        # Handle if already datetime objects
+        if isinstance(event_date, datetime):
+            event_date = event_date.date()
+        elif isinstance(event_date, str):
+            event_date = datetime.fromisoformat(event_date).date()
+
+        if isinstance(event_time, str):
+            event_time = datetime.strptime(event_time, "%H:%M:%S").time()
+        elif isinstance(event_time, datetime):
+            event_time = event_time.time()
+        elif isinstance(event_time, timedelta):
+            # Convert timedelta to time (MySQL TIME type returns timedelta)
+            total_seconds = int(event_time.total_seconds())
+            hours = total_seconds // 3600
+            minutes = (total_seconds % 3600) // 60
+            seconds = total_seconds % 60
+            event_time = time(hours, minutes, seconds)
+
+        # Combine date and time
+        return datetime.combine(event_date, event_time)
+    except Exception as e:
+        logger.warning(f"Failed to combine date {event_date} and time {event_time}: {e}")
+        return None
+
+
+class Consolidator:
+    """Consolidate MySQL rows into PostgreSQL format with node-based JSONB."""
+
+    @staticmethod
+    def consolidate_rawdatacor(mysql_rows: List[Dict[str, Any]]) -> Dict[str, Any]:
+        """Consolidate RAWDATACOR rows by NodeNum into single PostgreSQL row.
+
+        Args:
+            mysql_rows: List of MySQL rows with same (UnitName, ToolNameID, EventDate, EventTime)
+                       but different NodeNum
+
+        Returns:
+            Single PostgreSQL row dict with consolidated measurements
+
+        Example measurements structure:
+            {
+                "node_1": {
+                    "0": {"value": "123.45", "unit": "°C"},
+                    "1": {"value": "67.89", "unit": "bar"},
+                    ...
+                    "F": {"value": "11.22", "unit": "m/s"}
+                },
+                "node_2": { ... }
+            }
+        """
+        if not mysql_rows:
+            raise ValueError("Cannot consolidate empty row list")
+
+        # Get consolidation key from first row (all rows have same key)
+        first_row = mysql_rows[0]
+        config = TABLE_CONFIGS["RAWDATACOR"]
+
+        # Build measurements JSONB by node
+        measurements = {}
+        mysql_ids = []
+
+        for row in mysql_rows:
+            node_num = row.get("NodeNum")
+            if node_num is None:
+                logger.warning(f"Row missing NodeNum, skipping: {row.get(config['mysql_pk'])}")
+                continue
+
+            # Track MySQL IDs for max calculation
+            mysql_id = row.get(config["mysql_pk"])
+            if mysql_id:
+                mysql_ids.append(int(mysql_id))
+
+            # Create node key (e.g., "node_1", "node_2")
+            node_key = f"node_{node_num}"
+
+            # Build measurements for this node
+            node_measurements = {}
+            val_columns = RAWDATACOR_COLUMNS["val_columns"]
+            unit_columns = RAWDATACOR_COLUMNS["unit_columns"]
+
+            for idx, val_col in enumerate(val_columns):
+                unit_col = unit_columns[idx]
+                val = row.get(val_col)
+                unit = row.get(unit_col)
+
+                # Only include non-NULL values
+                if val is not None:
+                    # Get hex key (0, 1, 2, ..., F)
+                    hex_key = val_col.replace("Val", "")
+
+                    measurement = {}
+                    # Convert Decimal to float for JSON serialization
+                    if isinstance(val, Decimal):
+                        measurement["value"] = float(val)
+                    else:
+                        measurement["value"] = val
+
+                    if unit is not None and unit != "":
+                        measurement["unit"] = unit
+
+                    node_measurements[hex_key] = measurement
+
+            measurements[node_key] = node_measurements
+
+        # Calculate max MySQL ID
+        mysql_max_id = max(mysql_ids) if mysql_ids else 0
+
+        # Combine EventDate + EventTime into event_timestamp
+        event_timestamp = _combine_datetime(
+            first_row.get("EventDate"),
+            first_row.get("EventTime")
+        )
+
+        # Extract year from event_timestamp for partition key
+        event_year = event_timestamp.year if event_timestamp else None
+
+        # Calculate created_at (minimum from all rows)
+        # Note: RAWDATACOR does not have updated_at field
+        created_at_values = [row.get("created_at") for row in mysql_rows if row.get("created_at") is not None]
+        created_at = min(created_at_values) if created_at_values else None
+
+        # Build PostgreSQL row
+        pg_row = {
+            "mysql_max_id": mysql_max_id,
+            "unit_name": first_row.get("UnitName"),
+            "tool_name_id": first_row.get("ToolNameID"),
+            "event_timestamp": event_timestamp,
+            "event_year": event_year,
+            "measurements": measurements,
+            "bat_level": first_row.get("BatLevel"),
+            "temperature": first_row.get("Temperature"),
+            "bat_level_module": first_row.get("BatLevel_module"),
+            "temperature_module": first_row.get("Temperature_module"),
+            "rssi_module": first_row.get("RSSI_module"),
+            "created_at": created_at,
+        }
+
+        return pg_row
+
+    @staticmethod
+    def consolidate_elabdatadisp(mysql_rows: List[Dict[str, Any]]) -> Dict[str, Any]:
+        """Consolidate ELABDATADISP rows by NodeNum into single PostgreSQL row.
+
+        Args:
+            mysql_rows: List of MySQL rows with same (UnitName, ToolNameID, EventDate, EventTime)
+                       but different NodeNum
+
+        Returns:
+            Single PostgreSQL row dict with consolidated measurements
+
+        Example measurements structure:
+            {
+                "node_1": {
+                    "shifts": {"x": 1.234, "y": 2.345, "z": 3.456, ...},
+                    "coordinates": {"x": 10.123, "y": 20.234, ...},
+                    "kinematics": {"speed": 1.111, ...},
+                    "sensors": {"t_node": 25.5, ...},
+                    "calculated": {"alfa_x": 0.123, ...}
+                },
+                "node_2": { ... }
+            }
+        """
+        if not mysql_rows:
+            raise ValueError("Cannot consolidate empty row list")
+
+        # Get consolidation key from first row (all rows have same key)
+        first_row = mysql_rows[0]
+        config = TABLE_CONFIGS["ELABDATADISP"]
+
+        # Build measurements JSONB by node
+        measurements = {}
+        mysql_ids = []
+
+        for row in mysql_rows:
+            node_num = row.get("NodeNum")
+            if node_num is None:
+                logger.warning(f"Row missing NodeNum, skipping: {row.get(config['mysql_pk'])}")
+                continue
+
+            # Track MySQL IDs for max calculation
+            mysql_id = row.get(config["mysql_pk"])
+            if mysql_id:
+                mysql_ids.append(int(mysql_id))
+
+            # Create node key (e.g., "node_1", "node_2")
+            node_key = f"node_{node_num}"
+
+            # Build measurements for this node using field mapping
+            node_measurements = {
+                "shifts": {},
+                "coordinates": {},
+                "kinematics": {},
+                "sensors": {},
+                "calculated": {},
+            }
+
+            # Add state and calc_err if present
+            if "State" in row and row["State"] is not None:
+                node_measurements["state"] = row["State"]
+            # MySQL field is 'calcerr' (lowercase), not 'CalcErr'
+            if "calcerr" in row and row["calcerr"] is not None:
+                node_measurements["calc_err"] = row["calcerr"]
+
+            # Map MySQL fields to JSONB structure
+            for mysql_field, (category, json_key) in ELABDATADISP_FIELD_MAPPING.items():
+                value = row.get(mysql_field)
+
+                # Only include non-NULL values
+                if value is not None:
+                    # Convert Decimal to float for JSON serialization
+                    if isinstance(value, Decimal):
+                        value = float(value)
+
+                    node_measurements[category][json_key] = value
+
+            # Remove empty categories
+            node_measurements = {
+                k: v for k, v in node_measurements.items()
+                if v and (not isinstance(v, dict) or len(v) > 0)
+            }
+
+            measurements[node_key] = node_measurements
+
+        # Calculate max MySQL ID
+        mysql_max_id = max(mysql_ids) if mysql_ids else 0
+
+        # Combine EventDate + EventTime into event_timestamp
+        event_timestamp = _combine_datetime(
+            first_row.get("EventDate"),
+            first_row.get("EventTime")
+        )
+
+        # Extract year from event_timestamp for partition key
+        event_year = event_timestamp.year if event_timestamp else None
+
+        # Calculate created_at (minimum from all rows) and updated_at (maximum from all rows)
+        created_at_values = [row.get("created_at") for row in mysql_rows if row.get("created_at") is not None]
+        updated_at_values = [row.get("updated_at") for row in mysql_rows if row.get("updated_at") is not None]
+
+        created_at = min(created_at_values) if created_at_values else None
+        updated_at = max(updated_at_values) if updated_at_values else None
+
+        # Build PostgreSQL row
+        pg_row = {
+            "mysql_max_id": mysql_max_id,
+            "unit_name": first_row.get("UnitName"),
+            "tool_name_id": first_row.get("ToolNameID"),
+            "event_timestamp": event_timestamp,
+            "event_year": event_year,
+            "measurements": measurements,
+            "created_at": created_at,
+            "updated_at": updated_at,
+        }
+
+        return pg_row
+
+    @staticmethod
+    def consolidate(table: str, mysql_rows: List[Dict[str, Any]]) -> Dict[str, Any]:
+        """Consolidate rows for the specified table.
+
+        Args:
+            table: Table name (RAWDATACOR or ELABDATADISP)
+            mysql_rows: List of MySQL rows to consolidate
+
+        Returns:
+            Consolidated PostgreSQL row
+
+        Raises:
+            ValueError: If table is unknown or rows are empty
+        """
+        if not mysql_rows:
+            raise ValueError("Cannot consolidate empty row list")
+
+        table_upper = table.upper()
+
+        if table_upper == "RAWDATACOR":
+            return Consolidator.consolidate_rawdatacor(mysql_rows)
+        elif table_upper == "ELABDATADISP":
+            return Consolidator.consolidate_elabdatadisp(mysql_rows)
+        else:
+            raise ValueError(f"Unknown table: {table}")
+
+
+def consolidate_rows(table: str, mysql_rows: List[Dict[str, Any]]) -> Dict[str, Any]:
+    """Convenience function to consolidate rows.
+
+    Args:
+        table: Table name (RAWDATACOR or ELABDATADISP)
+        mysql_rows: List of MySQL rows with same consolidation key
+
+    Returns:
+        Single consolidated PostgreSQL row
+    """
+    return Consolidator.consolidate(table, mysql_rows)

src/migrator/full_migration.py

@@ -31,12 +31,13 @@ class FullMigrator:
         self.settings = get_settings()
         self.state = MigrationState()
 
-    def migrate(self, dry_run: bool = False, resume: bool = False) -> int:
+    def migrate(self, dry_run: bool = False, resume: bool = False, partition: str = None) -> int:
         """Perform full migration of the table with resume capability.
 
         Args:
             dry_run: If True, log what would be done but don't modify data
             resume: If True, resume from last checkpoint; if False, check for conflicts
+            partition: If specified, only migrate this partition (for testing/debugging)
 
         Returns:
             Total number of rows migrated in this run
@@ -70,21 +71,28 @@ class FullMigrator:
                     # - Full restart requires clearing the table
 
                     previous_migrated_count = self._get_previous_migrated_count(pg_conn, pg_table)
-                    last_completed_partition = self._get_last_completed_partition(pg_conn, pg_table)
+                    # If specific partition requested, ignore migration_state partition tracking
+                    # and start fresh for that partition
+                    if partition:
+                        last_completed_partition = None
+                        logger.info(f"Specific partition {partition} requested - ignoring migration_state partition tracking")
+                    else:
+                        last_completed_partition = self._get_last_completed_partition(pg_conn, pg_table)
 
                     if previous_migrated_count > 0:
                         pg_row_count = pg_conn.get_row_count(pg_table)
                         logger.warning(
                             f"Found previous migration state: {pg_row_count} rows already in {pg_table}"
                         )
-                        if not resume:
+                        if not resume and not dry_run:
                             raise ValueError(
                                 f"Migration already in progress for {pg_table}. "
                                 f"Use --resume to continue from last checkpoint, or delete data to restart."
                             )
-                        logger.info(f"Resuming migration - found {pg_row_count} existing rows")
-                        if last_completed_partition:
-                            logger.info(f"Last completed partition: {last_completed_partition}")
+                        if resume:
+                            logger.info(f"Resuming migration - found {pg_row_count} existing rows")
+                            if last_completed_partition:
+                                logger.info(f"Last completed partition: {last_completed_partition}")
                         # Progress bar tracks MySQL rows processed (before consolidation)
                         # Consolidation reduces count but not the rows we need to fetch
                         rows_to_migrate = total_rows
@@ -94,7 +102,10 @@ class FullMigrator:
                         rows_to_migrate = total_rows
 
                     if dry_run:
-                        logger.info(f"[DRY RUN] Would migrate {rows_to_migrate} rows")
+                        if partition:
+                            logger.info(f"[DRY RUN] Would migrate partition {partition} (~{rows_to_migrate} total rows in table)")
+                        else:
+                            logger.info(f"[DRY RUN] Would migrate {rows_to_migrate} rows")
                         return rows_to_migrate
 
                     migrated = previous_migrated_count
@@ -112,6 +123,13 @@ class FullMigrator:
                         partitions = mysql_conn.get_table_partitions(mysql_table)
                         logger.info(f"Found {len(partitions)} partitions for {mysql_table}")
 
+                        # If specific partition requested, filter to just that one
+                        if partition:
+                            if partition not in partitions:
+                                raise ValueError(f"Partition {partition} not found. Available: {partitions}")
+                            partitions = [partition]
+                            logger.info(f"Filtering to only partition: {partition}")
+
                         for partition_idx, partition in enumerate(partitions, 1):
                             # Skip partitions already completed in previous run
                             if last_completed_partition and partition <= last_completed_partition:
@@ -197,10 +215,13 @@ class FullMigrator:
                                         migrated += inserted
                                         batch_count += 1
                                         progress.update(fetched_in_buffer)
+                                        # Get the last inserted ID for resume capability
+                                        current_last_id = self._get_last_migrated_id(pg_conn, pg_table)
                                         # Update migration state after every batch flush
-                                        # Do NOT set last_completed_partition yet - partition is still being processed
+                                        # Keep last_completed_partition if we've completed partitions before
                                         self._update_migration_state(
-                                            pg_conn, migrated, None, migration_start_time
+                                            pg_conn, migrated, current_last_id, migration_start_time,
+                                            last_partition=last_processed_partition
                                         )
                                         logger.debug(
                                             f"Partition {partition}: flushed {inserted} rows, "
@@ -217,9 +238,12 @@ class FullMigrator:
                                     migrated += inserted
                                     batch_count += 1
                                     progress.update(fetched_in_buffer)
-                                    # Still don't set last_completed_partition - partition is still being finalized
+                                    # Get the last inserted ID for resume capability
+                                    current_last_id = self._get_last_migrated_id(pg_conn, pg_table)
+                                    # Keep last_completed_partition if we've completed partitions before
                                     self._update_migration_state(
-                                        pg_conn, migrated, None, migration_start_time
+                                        pg_conn, migrated, current_last_id, migration_start_time,
+                                        last_partition=last_processed_partition
                                     )
                                     logger.debug(
                                         f"Partition {partition} final flush: {inserted} rows, "
@@ -229,8 +253,10 @@ class FullMigrator:
                             # NOW partition is complete - update with completed partition
                             logger.info(f"Partition {partition} complete: {partition_group_count} groups consolidated")
                             last_processed_partition = partition  # Track this partition as processed
+                            # Get final last ID for this partition
+                            final_last_id = self._get_last_migrated_id(pg_conn, pg_table)
                             self._update_migration_state(
-                                pg_conn, migrated, None, migration_start_time,
+                                pg_conn, migrated, final_last_id, migration_start_time,
                                 last_partition=partition
                             )
 
@@ -409,7 +435,8 @@ class FullMigrator:
 def run_full_migration(
     table: str,
     dry_run: bool = False,
-    resume: bool = False
+    resume: bool = False,
+    partition: str = None
 ) -> int:
     """Run full migration for a table.
 
@@ -417,9 +444,10 @@ def run_full_migration(
         table: Table name to migrate
         dry_run: If True, show what would be done without modifying data
         resume: If True, resume from last checkpoint instead of starting fresh
+        partition: If specified, only migrate this partition (for testing/debugging)
 
     Returns:
         Number of rows migrated in this run
     """
     migrator = FullMigrator(table)
-    return migrator.migrate(dry_run=dry_run, resume=resume)
+    return migrator.migrate(dry_run=dry_run, resume=resume, partition=partition)

src/migrator/full_migrator.py

@@ -0,0 +1,240 @@
+"""Full migration orchestrator for MySQL to PostgreSQL.
+
+Coordinates migration of all partitions or a specific partition.
+"""
+from typing import Optional, List
+
+from config import get_settings, TABLE_CONFIGS
+from src.connectors.mysql_connector import MySQLConnector
+from src.connectors.postgres_connector import PostgreSQLConnector
+from src.migrator.partition_migrator import PartitionMigrator
+from src.migrator.state_manager import StateManager
+from src.utils.logger import get_logger
+
+logger = get_logger(__name__)
+
+
+class FullMigrator:
+    """Orchestrate full migration of all partitions."""
+
+    def __init__(self, table: str):
+        """Initialize full migrator.
+
+        Args:
+            table: Table name (RAWDATACOR or ELABDATADISP)
+        """
+        if table.upper() not in ("RAWDATACOR", "ELABDATADISP"):
+            raise ValueError(f"Unknown table: {table}")
+
+        self.table = table.upper()
+        self.config = TABLE_CONFIGS[self.table]
+        self.settings = get_settings()
+        self.partition_migrator = PartitionMigrator(table)
+
+    def migrate(
+        self,
+        partition: Optional[str] = None,
+        dry_run: bool = False,
+        resume: bool = False
+    ) -> int:
+        """Perform full migration.
+
+        Args:
+            partition: If specified, migrate only this partition (for testing)
+            dry_run: If True, log what would be done without modifying data
+            resume: If True, resume from last checkpoint
+
+        Returns:
+            Total number of PostgreSQL rows migrated
+        """
+        mysql_table = self.config["mysql_table"]
+        pg_table = self.config["postgres_table"]
+
+        logger.info(f"Starting full migration of {mysql_table} -> {pg_table}")
+
+        try:
+            with MySQLConnector() as mysql_conn:
+                with PostgreSQLConnector() as pg_conn:
+                    # Check PostgreSQL table exists
+                    if not pg_conn.table_exists(pg_table):
+                        raise ValueError(
+                            f"PostgreSQL table {pg_table} does not exist. "
+                            "Run 'setup --create-schema' first."
+                        )
+
+                    # Initialize state manager (using _global for table-level state)
+                    global_state_mgr = StateManager(pg_conn, pg_table, "_global")
+
+                    # Get partitions to migrate
+                    if partition:
+                        # Single partition mode
+                        partitions = [partition]
+                        # Check if this partition is in-progress and should be resumed
+                        in_progress_partitions = global_state_mgr.get_in_progress_partitions()
+                        should_resume = resume and partition in in_progress_partitions
+                        logger.info(f"Single partition mode: {partition}" +
+                                   (f" (resuming from checkpoint)" if should_resume else ""))
+                    else:
+                        # Full migration: all partitions
+                        all_partitions = mysql_conn.get_table_partitions(mysql_table)
+                        logger.info(f"Found {len(all_partitions)} partitions: {all_partitions}")
+
+                        # Check resume - get list of completed and in-progress partitions
+                        if resume:
+                            completed_partitions = global_state_mgr.get_completed_partitions()
+                            in_progress_partitions = global_state_mgr.get_in_progress_partitions()
+
+                            # Skip partitions already completed
+                            partitions = [p for p in all_partitions if p not in completed_partitions]
+
+                            logger.info(
+                                f"Resuming migration. Completed: {len(completed_partitions)}, "
+                                f"In-progress: {len(in_progress_partitions)}, "
+                                f"Remaining: {len(partitions)} partitions"
+                            )
+
+                            if in_progress_partitions:
+                                logger.info(f"Will resume in-progress partitions: {in_progress_partitions}")
+                        else:
+                            partitions = all_partitions
+                            in_progress_partitions = []
+
+                    if dry_run:
+                        logger.info(
+                            f"[DRY RUN] Would migrate {len(partitions)} partition(s): "
+                            f"{partitions}"
+                        )
+                        return 0
+
+                    # Migrate each partition
+                    total_migrated = 0
+                    for idx, part_name in enumerate(partitions, 1):
+                        # Determine if this partition should be resumed
+                        # In single partition mode, use should_resume from above
+                        # In full mode, check if partition is in-progress
+                        if partition:
+                            # Single partition mode - use the should_resume flag set earlier
+                            do_resume = should_resume
+                        else:
+                            # Full migration mode - resume if partition is in-progress
+                            do_resume = resume and part_name in in_progress_partitions
+
+                        resume_msg = " (resuming from checkpoint)" if do_resume else ""
+                        logger.info(
+                            f"[{idx}/{len(partitions)}] Migrating partition: {part_name}{resume_msg}"
+                        )
+
+                        try:
+                            migrated = self.partition_migrator.migrate_partition(
+                                mysql_conn,
+                                pg_conn,
+                                part_name,
+                                dry_run=dry_run,
+                                resume=do_resume
+                            )
+                            total_migrated += migrated
+                            logger.info(
+                                f"✓ Partition {part_name} complete: "
+                                f"{migrated} rows, total: {total_migrated}"
+                            )
+
+                        except Exception as e:
+                            logger.error(
+                                f"Failed to migrate partition {part_name}: {e}"
+                            )
+                            raise
+
+                    # Get final row count from PostgreSQL
+                    final_count = pg_conn.get_row_count(pg_table)
+
+                    # Only mark global migration complete if ALL partitions are completed
+                    # Check if there are any in-progress or pending partitions remaining
+                    if partition:
+                        # Single partition mode - don't update global state
+                        logger.info(
+                            f"✓ Partition migration complete: {total_migrated:,} rows migrated in this run, "
+                            f"{final_count:,} total rows in {pg_table}"
+                        )
+                    else:
+                        # Full migration mode - check if everything is really done
+                        completed_partitions = global_state_mgr.get_completed_partitions()
+                        in_progress_check = global_state_mgr.get_in_progress_partitions()
+                        all_partitions_check = mysql_conn.get_table_partitions(mysql_table)
+
+                        if len(completed_partitions) == len(all_partitions_check) and len(in_progress_check) == 0:
+                            # All partitions are completed - mark global as complete
+                            # Get the maximum last_key across all partitions for incremental migration
+                            max_last_key = global_state_mgr.get_max_last_key_across_partitions()
+                            if max_last_key:
+                                logger.info(
+                                    f"Setting global last_key: "
+                                    f"({max_last_key.get('unit_name')}, {max_last_key.get('tool_name_id')}, "
+                                    f"{max_last_key.get('event_date')}, {max_last_key.get('event_time')})"
+                                )
+                                global_state_mgr.update_state(last_key=max_last_key)
+
+                            global_state_mgr.mark_completed(final_count)
+                            logger.info(
+                                f"✓ Full migration complete: {total_migrated:,} rows migrated in this run, "
+                                f"{final_count:,} total rows in {pg_table}"
+                            )
+                        else:
+                            logger.info(
+                                f"✓ Migration batch complete: {total_migrated:,} rows migrated in this run, "
+                                f"{final_count:,} total rows in {pg_table}. "
+                                f"{len(completed_partitions)}/{len(all_partitions_check)} partitions completed."
+                            )
+
+                    return total_migrated
+
+        except Exception as e:
+            logger.error(f"Full migration failed: {e}")
+            raise
+
+    def _get_remaining_partitions(
+        self,
+        all_partitions: List[str],
+        last_completed: str
+    ) -> List[str]:
+        """Get list of partitions that still need to be migrated.
+
+        Args:
+            all_partitions: All partition names from MySQL
+            last_completed: Last partition that was completed
+
+        Returns:
+            List of partition names to migrate
+        """
+        try:
+            # Find index of last completed partition
+            idx = all_partitions.index(last_completed)
+            # Return partitions after it
+            return all_partitions[idx + 1:]
+        except ValueError:
+            # last_completed not in list, return all
+            logger.warning(
+                f"Last completed partition {last_completed} not found in partition list. "
+                "Starting from beginning."
+            )
+            return all_partitions
+
+
+def run_full_migration(
+    table: str,
+    partition: Optional[str] = None,
+    dry_run: bool = False,
+    resume: bool = False
+) -> int:
+    """Run full migration for a table.
+
+    Args:
+        table: Table name to migrate (RAWDATACOR or ELABDATADISP)
+        partition: If specified, migrate only this partition
+        dry_run: If True, show what would be done without modifying data
+        resume: If True, resume from last checkpoint
+
+    Returns:
+        Number of rows migrated
+    """
+    migrator = FullMigrator(table)
+    return migrator.migrate(partition=partition, dry_run=dry_run, resume=resume)

src/migrator/incremental_migration.py

@@ -158,6 +158,8 @@ class IncrementalMigrator:
 
         # Fetch and migrate rows in batches
         batch_count = 0
+        progress = None
+
         for batch in mysql_conn.fetch_rows_since(
             mysql_table,
             last_timestamp
@@ -172,11 +174,23 @@ class IncrementalMigrator:
                 )
                 progress.__enter__()
 
-            # Transform batch
-            transformed = DataTransformer.transform_batch(
-                mysql_table,
-                batch
-            )
+            # For tables that need consolidation, group rows by consolidation key
+            # and consolidate before transforming
+            if mysql_table in ("RAWDATACOR", "ELABDATADISP"):
+                # Group batch by consolidation key
+                consolidated_batch = self._consolidate_batch(batch, mysql_table)
+                # Transform consolidated batch
+                transformed = DataTransformer.transform_batch(
+                    mysql_table,
+                    consolidated_batch,
+                    consolidate=False  # Already consolidated above
+                )
+            else:
+                # No consolidation needed for other tables
+                transformed = DataTransformer.transform_batch(
+                    mysql_table,
+                    batch
+                )
 
             # Insert batch
             columns = DataTransformer.get_column_order(pg_table)
@@ -187,7 +201,8 @@ class IncrementalMigrator:
             )
 
             migrated += inserted
-            progress.update(inserted)
+            if progress:
+                progress.update(inserted)
 
         if batch_count == 0:
             logger.info(f"No new rows to migrate for {mysql_table}")
@@ -267,11 +282,23 @@ class IncrementalMigrator:
                 if not batch:
                     break
 
-                # Transform batch
-                transformed = DataTransformer.transform_batch(
-                    mysql_table,
-                    batch
-                )
+                # For tables that need consolidation, group rows by consolidation key
+                # and consolidate before transforming
+                if mysql_table in ("RAWDATACOR", "ELABDATADISP"):
+                    # Group batch by consolidation key
+                    consolidated_batch = self._consolidate_batch(batch, mysql_table)
+                    # Transform consolidated batch
+                    transformed = DataTransformer.transform_batch(
+                        mysql_table,
+                        consolidated_batch,
+                        consolidate=False  # Already consolidated above
+                    )
+                else:
+                    # No consolidation needed for other tables
+                    transformed = DataTransformer.transform_batch(
+                        mysql_table,
+                        batch
+                    )
 
                 # Insert batch
                 columns = DataTransformer.get_column_order(pg_table)
@@ -304,6 +331,59 @@ class IncrementalMigrator:
 
         return migrated
 
+    def _consolidate_batch(self, batch: list, mysql_table: str) -> list:
+        """Consolidate batch by grouping rows with same consolidation key.
+
+        Args:
+            batch: List of rows from MySQL
+            mysql_table: Table name (RAWDATACOR or ELABDATADISP)
+
+        Returns:
+            List of rows, already consolidated by consolidation key
+        """
+        if not batch:
+            return batch
+
+        # Group rows by consolidation key
+        groups = {}
+        for row in batch:
+            # Build consolidation key
+            if mysql_table == "ELABDATADISP":
+                key = (
+                    row.get("UnitName"),
+                    row.get("ToolNameID"),
+                    row.get("EventDate"),
+                    row.get("EventTime")
+                )
+            else:  # RAWDATACOR
+                key = (
+                    row.get("UnitName"),
+                    row.get("ToolNameID"),
+                    row.get("EventDate"),
+                    row.get("EventTime")
+                )
+
+            if key not in groups:
+                groups[key] = []
+            groups[key].append(row)
+
+        # Consolidate each group and flatten back to list
+        consolidated = []
+        for key, group_rows in groups.items():
+            if len(group_rows) == 1:
+                # Single row, no consolidation needed
+                consolidated.append(group_rows[0])
+            else:
+                # Multiple rows with same key, consolidate them
+                consolidated_group = DataTransformer.consolidate_elabdatadisp_batch(
+                    group_rows
+                ) if mysql_table == "ELABDATADISP" else DataTransformer.consolidate_rawdatacor_batch(
+                    group_rows
+                )
+                consolidated.extend(consolidated_group)
+
+        return consolidated
+
 
 def run_incremental_migration(
     table: str,

src/migrator/incremental_migrator.py

@@ -0,0 +1,383 @@
+"""Incremental migration from MySQL to PostgreSQL.
+
+Migrates only new data since last full/incremental migration based on consolidation keys.
+"""
+from typing import Optional
+
+from config import get_settings, TABLE_CONFIGS
+from src.connectors.mysql_connector import MySQLConnector
+from src.connectors.postgres_connector import PostgreSQLConnector
+from src.migrator.consolidator import consolidate_rows
+from src.migrator.state_manager import StateManager
+from src.utils.logger import get_logger
+from src.utils.progress import ProgressTracker
+from src.utils.validation import validate_consolidation_key, ErrorLogger
+
+logger = get_logger(__name__)
+
+
+class IncrementalMigrator:
+    """Perform incremental migration based on consolidation keys."""
+
+    def __init__(self, table: str):
+        """Initialize incremental migrator.
+
+        Args:
+            table: Table name (RAWDATACOR or ELABDATADISP)
+        """
+        if table.upper() not in ("RAWDATACOR", "ELABDATADISP"):
+            raise ValueError(f"Unknown table: {table}")
+
+        self.table = table.upper()
+        self.config = TABLE_CONFIGS[self.table]
+        self.settings = get_settings()
+
+    def migrate(self, dry_run: bool = False) -> int:
+        """Perform incremental migration since last migration.
+
+        Migrates only consolidation keys that come AFTER the last migrated key.
+
+        Args:
+            dry_run: If True, log what would be done without modifying data
+
+        Returns:
+            Number of PostgreSQL rows migrated
+        """
+        mysql_table = self.config["mysql_table"]
+        pg_table = self.config["postgres_table"]
+
+        logger.info(f"Starting incremental migration of {mysql_table} -> {pg_table}")
+
+        try:
+            with MySQLConnector() as mysql_conn:
+                with PostgreSQLConnector() as pg_conn:
+                    # Check PostgreSQL table exists
+                    if not pg_conn.table_exists(pg_table):
+                        raise ValueError(
+                            f"PostgreSQL table {pg_table} does not exist. "
+                            "Run 'setup --create-schema' first."
+                        )
+
+                    # Initialize state manager
+                    state_mgr = StateManager(pg_conn, pg_table)
+
+                    # Initialize error logger
+                    error_logger = ErrorLogger(pg_table, "incremental")
+
+                    # Get last migrated key from migration_state
+                    # This was saved during the last full/incremental migration
+                    last_key = state_mgr.get_last_key()
+
+                    if last_key is None:
+                        logger.info(
+                            f"No previous migration found for {pg_table}. "
+                            "Run 'migrate full' for initial migration."
+                        )
+                        return 0
+
+                    logger.info(
+                        f"Last migrated key (from migration_state): "
+                        f"({last_key.get('unit_name')}, {last_key.get('tool_name_id')}, "
+                        f"{last_key.get('event_date')}, {last_key.get('event_time')})"
+                    )
+
+                    # Determine which partitions to process based on last_key's event_date
+                    # This is a MAJOR optimization: instead of querying the entire table,
+                    # we only process partitions from the last migrated year onwards
+                    from config import get_partitions_from_year, date_string_to_partition_name
+
+                    last_event_date = last_key.get('event_date')
+                    if not last_event_date:
+                        logger.warning("Last key has no event_date, starting from 2014")
+                        partitions_to_process = mysql_conn.get_table_partitions(mysql_table)
+                    else:
+                        # Extract year from last_event_date and get partitions from that year onwards
+                        year = int(str(last_event_date)[:4]) if len(str(last_event_date)) >= 4 else 2014
+                        partitions_to_process = get_partitions_from_year(year, mysql_table)
+
+                        logger.info(
+                            f"Optimized incremental sync: Processing only {len(partitions_to_process)} "
+                            f"partitions from year {year} onwards: {partitions_to_process}"
+                        )
+                        logger.info(
+                            f"Skipping partitions before year {year} (no new data possible there)"
+                        )
+
+                    if dry_run:
+                        # In dry-run, check how many new keys exist in MySQL
+                        logger.info("[DRY RUN] Checking for new keys across relevant partitions...")
+
+                        total_new_keys = 0
+                        first_keys_found = []
+
+                        for partition in partitions_to_process:
+                            # For first partition (same year as last_key), use after_key
+                            # For subsequent partitions, start from beginning
+                            if partition == partitions_to_process[0]:
+                                sample_keys = mysql_conn.fetch_consolidation_keys_from_partition_after(
+                                    mysql_table,
+                                    partition=partition,
+                                    after_key=last_key,
+                                    limit=100
+                                )
+                                logger.info(f"[DRY RUN] Partition {partition}: {len(sample_keys)} new keys (after last_key)")
+                            else:
+                                sample_keys = mysql_conn.fetch_consolidation_keys_from_partition_after(
+                                    mysql_table,
+                                    partition=partition,
+                                    after_key=None,  # All keys from this partition are new
+                                    limit=100
+                                )
+                                logger.info(f"[DRY RUN] Partition {partition}: {len(sample_keys)} new keys (all new)")
+
+                            total_new_keys += len(sample_keys)
+                            if sample_keys and len(first_keys_found) < 3:
+                                first_keys_found.extend(sample_keys[:3 - len(first_keys_found)])
+
+                        if total_new_keys > 0:
+                            logger.info(f"[DRY RUN] Found at least {total_new_keys} new keys across {len(partitions_to_process)} partitions")
+                            logger.info("[DRY RUN] First 3 keys:")
+                            for i, key in enumerate(first_keys_found[:3]):
+                                logger.info(
+                                    f"  {i+1}. ({key.get('UnitName')}, {key.get('ToolNameID')}, "
+                                    f"{key.get('EventDate')}, {key.get('EventTime')})"
+                                )
+                            logger.info(
+                                f"[DRY RUN] Run without --dry-run to perform actual migration"
+                            )
+                            return total_new_keys
+                        else:
+                            logger.info("[DRY RUN] No new keys found - database is up to date")
+                            return 0
+
+                    # Migrate new keys
+                    migrated_rows = 0
+                    insert_buffer = []
+                    buffer_size = self.settings.migration.consolidation_group_limit // 10
+                    last_processed_key = None  # Track last key for final state update
+
+                    with ProgressTracker(
+                        total=None,  # Unknown total
+                        description=f"Incremental migration of {mysql_table}"
+                    ) as progress:
+
+                        # Get column order for PostgreSQL insert
+                        pg_columns = self._get_pg_columns()
+
+                        # Process each partition
+                        for partition_idx, partition in enumerate(partitions_to_process, 1):
+                            logger.info(
+                                f"[{partition_idx}/{len(partitions_to_process)}] "
+                                f"Processing partition {partition}..."
+                            )
+
+                            # For first partition (same year as last_key), fetch keys AFTER last_key
+                            # For subsequent partitions, fetch ALL keys (they're all new)
+                            use_after_key = last_key if partition == partitions_to_process[0] else None
+
+                            while True:
+                                # Fetch batch of consolidation keys from this partition
+                                keys = mysql_conn.fetch_consolidation_keys_from_partition_after(
+                                    mysql_table,
+                                    partition=partition,
+                                    after_key=use_after_key,
+                                    limit=self.settings.migration.consolidation_group_limit
+                                )
+
+                                if not keys:
+                                    logger.info(f"  No more keys in partition {partition}")
+                                    break
+
+                                logger.info(f"  Processing {len(keys)} keys from {partition}...")
+
+                                # Process each consolidation key
+                                keys_processed = 0
+                                for key in keys:
+                                    keys_processed += 1
+                                    # Log progress every 1000 keys
+                                    if keys_processed % 1000 == 0:
+                                        logger.info(f"  Processed {keys_processed}/{len(keys)} keys in this batch...")
+                                    unit_name = key.get("UnitName")
+                                    tool_name_id = key.get("ToolNameID")
+                                    event_date = key.get("EventDate")
+                                    event_time = key.get("EventTime")
+
+                                    # Validate consolidation key before fetching
+                                    is_valid, error_reason = validate_consolidation_key(
+                                        unit_name, tool_name_id, event_date, event_time
+                                    )
+
+                                    if not is_valid:
+                                        # Log invalid key and skip
+                                        error_logger.log_invalid_key(
+                                            unit_name, tool_name_id, event_date, event_time, error_reason
+                                        )
+                                        continue
+
+                                    # Fetch all MySQL rows for this key (all nodes, all partitions)
+                                    try:
+                                        mysql_rows = mysql_conn.fetch_records_for_key_all_partitions(
+                                            mysql_table,
+                                            unit_name,
+                                            tool_name_id,
+                                            event_date,
+                                            event_time
+                                        )
+                                    except Exception as e:
+                                        # Log corrupted key that caused fetch error
+                                        error_logger.log_invalid_key(
+                                            unit_name, tool_name_id, event_date, event_time,
+                                            f"Fetch failed: {e}"
+                                        )
+                                        continue
+
+                                    if not mysql_rows:
+                                        logger.warning(
+                                            f"No records found for key: "
+                                            f"({unit_name}, {tool_name_id}, {event_date}, {event_time})"
+                                        )
+                                        continue
+
+                                    # Consolidate into single PostgreSQL row
+                                    try:
+                                        pg_row = consolidate_rows(self.table, mysql_rows)
+                                    except Exception as e:
+                                        logger.error(
+                                            f"Failed to consolidate key "
+                                            f"({unit_name}, {tool_name_id}, {event_date}, {event_time}): {e}"
+                                        )
+                                        continue
+
+                                    # Add to insert buffer
+                                    insert_buffer.append(pg_row)
+
+                                    # Track last processed key
+                                    last_processed_key = {
+                                        "unit_name": unit_name,
+                                        "tool_name_id": tool_name_id,
+                                        "event_date": str(event_date) if event_date else None,
+                                        "event_time": str(event_time) if event_time else None,
+                                    }
+
+                                    # Flush buffer when full
+                                    if len(insert_buffer) >= buffer_size:
+                                        # Use COPY with ON CONFLICT to handle duplicates
+                                        inserted = pg_conn.copy_from_with_conflict(
+                                            pg_table,
+                                            insert_buffer,
+                                            pg_columns,
+                                            conflict_columns=["unit_name", "tool_name_id", "event_timestamp", "event_year"]
+                                        )
+                                        migrated_rows += inserted
+                                        progress.update(inserted)
+
+                                        # Update state with last key (from tracked variable)
+                                        state_mgr.update_state(
+                                            last_key=last_processed_key,
+                                            total_rows_migrated=state_mgr.get_total_rows_migrated() + migrated_rows
+                                        )
+
+                                        logger.debug(
+                                            f"Flushed {inserted} rows, total new: {migrated_rows}"
+                                        )
+                                        insert_buffer = []
+
+                                # After processing all keys in batch, update use_after_key for next iteration
+                                if keys:
+                                    last_key_in_batch = keys[-1]
+                                    use_after_key = {
+                                        "unit_name": last_key_in_batch.get("UnitName"),
+                                        "tool_name_id": last_key_in_batch.get("ToolNameID"),
+                                        "event_date": str(last_key_in_batch.get("EventDate")) if last_key_in_batch.get("EventDate") else None,
+                                        "event_time": str(last_key_in_batch.get("EventTime")) if last_key_in_batch.get("EventTime") else None,
+                                    }
+
+                        # Flush remaining buffer
+                        if insert_buffer:
+                            # Use COPY with ON CONFLICT to handle duplicates
+                            inserted = pg_conn.copy_from_with_conflict(
+                                pg_table,
+                                insert_buffer,
+                                pg_columns,
+                                conflict_columns=["unit_name", "tool_name_id", "event_timestamp", "event_year"]
+                            )
+                            migrated_rows += inserted
+                            progress.update(inserted)
+                            logger.debug(f"Final flush: {inserted} rows")
+
+                            # Update state with last key after final flush
+                            if last_processed_key:
+                                state_mgr.update_state(
+                                    last_key=last_processed_key,
+                                    total_rows_migrated=state_mgr.get_total_rows_migrated() + migrated_rows
+                                )
+
+                    # Get final row count
+                    final_count = pg_conn.get_row_count(pg_table)
+                    logger.info(f"Total PostgreSQL rows: {final_count}")
+
+                    # Close error logger and get count
+                    error_logger.close()
+
+                    logger.info(
+                        f"✓ Incremental migration complete: "
+                        f"{migrated_rows} new rows migrated to {pg_table}, "
+                        f"{error_logger.get_error_count()} invalid keys skipped"
+                    )
+
+                    return migrated_rows
+
+        except Exception as e:
+            logger.error(f"Incremental migration failed: {e}")
+            raise
+
+    def _get_pg_columns(self) -> list[str]:
+        """Get PostgreSQL column names in insertion order.
+
+        Returns:
+            List of column names
+        """
+        # Base columns for both tables
+        columns = [
+            "mysql_max_id",
+            "unit_name",
+            "tool_name_id",
+            "event_timestamp",
+            "event_year",
+            "measurements",
+        ]
+
+        # Add table-specific columns
+        if self.table == "RAWDATACOR":
+            columns.extend([
+                "bat_level",
+                "temperature",
+                "bat_level_module",
+                "temperature_module",
+                "rssi_module",
+                "created_at",
+            ])
+        elif self.table == "ELABDATADISP":
+            columns.extend([
+                "created_at",
+                "updated_at",
+            ])
+
+        return columns
+
+
+def run_incremental_migration(
+    table: str,
+    dry_run: bool = False
+) -> int:
+    """Run incremental migration for a table.
+
+    Args:
+        table: Table name to migrate (RAWDATACOR or ELABDATADISP)
+        dry_run: If True, show what would be done without modifying data
+
+    Returns:
+        Number of rows migrated
+    """
+    migrator = IncrementalMigrator(table)
+    return migrator.migrate(dry_run=dry_run)

src/migrator/parallel_migrator.py

@@ -0,0 +1,220 @@
+"""Parallel migration orchestrator for MySQL to PostgreSQL.
+
+Runs multiple partition migrations in parallel using multiprocessing.
+Each process migrates a different partition independently.
+"""
+import multiprocessing as mp
+from typing import Optional, List
+import sys
+
+from config import get_settings, TABLE_CONFIGS
+from src.connectors.mysql_connector import MySQLConnector
+from src.connectors.postgres_connector import PostgreSQLConnector
+from src.migrator.partition_migrator import PartitionMigrator
+from src.migrator.state_manager import StateManager
+from src.utils.logger import get_logger
+
+logger = get_logger(__name__)
+
+
+def migrate_partition_worker(table: str, partition_name: str, dry_run: bool = False, resume: bool = False) -> tuple[str, int, bool]:
+    """Worker function to migrate a single partition.
+
+    Runs in a separate process.
+
+    Args:
+        table: Table name (RAWDATACOR or ELABDATADISP)
+        partition_name: Partition to migrate
+        dry_run: If True, simulate without writing
+        resume: If True, resume from last checkpoint if exists
+
+    Returns:
+        Tuple of (partition_name, rows_migrated, success)
+    """
+    # Configure logging for this worker process (multiprocessing requires per-process setup)
+    # Set up root logger so all child loggers (PartitionMigrator, ProgressTracker, etc.) work
+    from src.utils.logger import setup_logger
+    setup_logger("")  # Configure root logger for this process
+
+    # Now get a logger for this worker
+    from src.utils.logger import get_logger
+    worker_logger = get_logger(f"worker.{partition_name}")
+
+    try:
+        resume_msg = " (resuming)" if resume else ""
+        worker_logger.info(f"Worker starting: {partition_name}{resume_msg}")
+
+        with MySQLConnector() as mysql_conn:
+            with PostgreSQLConnector() as pg_conn:
+                migrator = PartitionMigrator(table)
+                rows = migrator.migrate_partition(
+                    mysql_conn,
+                    pg_conn,
+                    partition_name,
+                    dry_run=dry_run,
+                    resume=resume
+                )
+
+                worker_logger.info(f"Worker completed: {partition_name} ({rows} rows)")
+                return (partition_name, rows, True)
+
+    except Exception as e:
+        worker_logger.error(f"Worker failed for {partition_name}: {e}")
+        return (partition_name, 0, False)
+
+
+class ParallelMigrator:
+    """Orchestrate parallel migration of multiple partitions."""
+
+    def __init__(self, table: str):
+        """Initialize parallel migrator.
+
+        Args:
+            table: Table name (RAWDATACOR or ELABDATADISP)
+        """
+        if table.upper() not in ("RAWDATACOR", "ELABDATADISP"):
+            raise ValueError(f"Unknown table: {table}")
+
+        self.table = table.upper()
+        self.config = TABLE_CONFIGS[self.table]
+        self.settings = get_settings()
+
+    def migrate(
+        self,
+        num_workers: int = 5,
+        dry_run: bool = False,
+        resume: bool = False
+    ) -> int:
+        """Perform parallel migration of all partitions.
+
+        Args:
+            num_workers: Number of parallel workers
+            dry_run: If True, log what would be done without modifying data
+            resume: If True, skip already completed partitions
+
+        Returns:
+            Total number of PostgreSQL rows migrated
+        """
+        mysql_table = self.config["mysql_table"]
+        pg_table = self.config["postgres_table"]
+
+        logger.info(
+            f"Starting parallel migration of {mysql_table} -> {pg_table} "
+            f"with {num_workers} workers"
+        )
+
+        try:
+            with MySQLConnector() as mysql_conn:
+                with PostgreSQLConnector() as pg_conn:
+                    # Check PostgreSQL table exists
+                    if not pg_conn.table_exists(pg_table):
+                        raise ValueError(
+                            f"PostgreSQL table {pg_table} does not exist. "
+                            "Run 'setup --create-schema' first."
+                        )
+
+                    # Get all partitions
+                    all_partitions = mysql_conn.get_table_partitions(mysql_table)
+                    logger.info(f"Found {len(all_partitions)} partitions: {all_partitions}")
+
+                    # Filter out completed partitions if resuming
+                    partitions_to_migrate = all_partitions
+                    if resume:
+                        state_mgr = StateManager(pg_conn, pg_table, "_global")
+                        completed = state_mgr.get_completed_partitions()
+                        if completed:
+                            partitions_to_migrate = [p for p in all_partitions if p not in completed]
+                            logger.info(
+                                f"Resuming: {len(completed)} partitions already completed, "
+                                f"{len(partitions_to_migrate)} remaining"
+                            )
+
+                    if not partitions_to_migrate:
+                        logger.info("All partitions already migrated")
+                        return 0
+
+                    if dry_run:
+                        logger.info(
+                            f"[DRY RUN] Would migrate {len(partitions_to_migrate)} partitions "
+                            f"in parallel with {num_workers} workers"
+                        )
+                        return 0
+
+            # Run migrations in parallel using multiprocessing pool
+            logger.info(f"Launching {num_workers} worker processes...")
+
+            with mp.Pool(processes=num_workers) as pool:
+                # Create tasks for each partition
+                tasks = [
+                    (self.table, partition_name, dry_run, resume)
+                    for partition_name in partitions_to_migrate
+                ]
+
+                # Run migrations in parallel
+                results = pool.starmap(migrate_partition_worker, tasks)
+
+            # Collect results
+            total_migrated = 0
+            failed_partitions = []
+
+            for partition_name, rows, success in results:
+                if success:
+                    total_migrated += rows
+                    logger.info(f"✓ {partition_name}: {rows} rows")
+                else:
+                    failed_partitions.append(partition_name)
+                    logger.error(f"✗ {partition_name}: FAILED")
+
+            if failed_partitions:
+                raise Exception(
+                    f"Migration failed for {len(failed_partitions)} partitions: "
+                    f"{', '.join(failed_partitions)}"
+                )
+
+            # Only mark global migration complete if ALL partitions are completed
+            with PostgreSQLConnector() as pg_conn:
+                final_count = pg_conn.get_row_count(pg_table)
+                global_state_mgr = StateManager(pg_conn, pg_table, "_global")
+
+                # Verify all partitions are actually completed
+                completed = global_state_mgr.get_completed_partitions()
+                in_progress = global_state_mgr.get_in_progress_partitions()
+
+                if len(in_progress) == 0:
+                    # All partitions completed successfully
+                    global_state_mgr.mark_completed(final_count)
+                    logger.info(
+                        f"✓ Parallel migration complete: {final_count} total rows in {pg_table}"
+                    )
+                else:
+                    logger.warning(
+                        f"Migration finished but {len(in_progress)} partitions still in-progress: {in_progress}. "
+                        f"Not marking global as completed."
+                    )
+
+            return final_count
+
+        except Exception as e:
+            logger.error(f"Parallel migration failed: {e}")
+            raise
+
+
+def run_parallel_migration(
+    table: str,
+    num_workers: int = 5,
+    dry_run: bool = False,
+    resume: bool = False
+) -> int:
+    """Run parallel migration for a table.
+
+    Args:
+        table: Table name to migrate (RAWDATACOR or ELABDATADISP)
+        num_workers: Number of parallel workers
+        dry_run: If True, show what would be done without modifying data
+        resume: If True, skip already completed partitions
+
+    Returns:
+        Number of rows migrated
+    """
+    migrator = ParallelMigrator(table)
+    return migrator.migrate(num_workers=num_workers, dry_run=dry_run, resume=resume)

src/migrator/partition_migrator.py

@@ -0,0 +1,358 @@
+"""Partition-based migration for MySQL to PostgreSQL.
+
+Streaming Strategy:
+1. Stream all rows from partition ordered by consolidation key (single query)
+2. Group rows by consolidation key as they arrive
+3. Validate and consolidate each group
+4. Batch insert using PostgreSQL COPY (10-100x faster than INSERT)
+5. Log invalid keys to error file
+6. Update migration state periodically with resume support
+"""
+from typing import Optional, List, Dict, Any
+from datetime import datetime
+
+from config import get_settings, TABLE_CONFIGS
+from src.connectors.mysql_connector import MySQLConnector
+from src.connectors.postgres_connector import PostgreSQLConnector
+from src.migrator.consolidator import consolidate_rows
+from src.migrator.state_manager import StateManager
+from src.utils.logger import get_logger
+from src.utils.progress import ProgressTracker
+from src.utils.validation import validate_consolidation_key, ErrorLogger
+
+logger = get_logger(__name__)
+
+
+class PartitionMigrator:
+    """Migrate a single MySQL partition to PostgreSQL with consolidation."""
+
+    def __init__(self, table: str):
+        """Initialize partition migrator.
+
+        Args:
+            table: Table name (RAWDATACOR or ELABDATADISP)
+        """
+        if table.upper() not in ("RAWDATACOR", "ELABDATADISP"):
+            raise ValueError(f"Unknown table: {table}")
+
+        self.table = table.upper()
+        self.config = TABLE_CONFIGS[self.table]
+        self.settings = get_settings()
+
+    def migrate_partition_streaming(
+        self,
+        mysql_conn: MySQLConnector,
+        pg_conn: PostgreSQLConnector,
+        partition_name: str,
+        dry_run: bool = False,
+        resume: bool = False
+    ) -> int:
+        """Migrate a partition using streaming (OPTIMIZED - eliminates N+1 queries).
+
+        This method streams all rows from the partition in a single query,
+        groups them by consolidation key, and uses PostgreSQL COPY for fast inserts.
+
+        Performance improvement: 10-100x faster than old N+1 query approach.
+
+        Args:
+            mysql_conn: MySQL connector (already connected)
+            pg_conn: PostgreSQL connector (already connected)
+            partition_name: MySQL partition name (e.g., 'p2024')
+            dry_run: If True, log what would be done without modifying data
+            resume: If True, resume from last checkpoint
+
+        Returns:
+            Number of PostgreSQL rows inserted (consolidated)
+        """
+        mysql_table = self.config["mysql_table"]
+        pg_table = self.config["postgres_table"]
+
+        logger.info(f"Starting STREAMING migration of partition {partition_name} from {mysql_table}")
+
+        # Initialize state manager for this partition
+        state_mgr = StateManager(pg_conn, pg_table, partition_name)
+
+        # Initialize error logger
+        error_logger = ErrorLogger(pg_table, partition_name)
+
+        # Check resume state
+        resume_from_key = None
+        start_key_tuple = None
+        if resume:
+            last_key = state_mgr.get_last_key()
+            if last_key:
+                resume_from_key = last_key
+                # Convert to tuple for MySQL query
+                start_key_tuple = (
+                    last_key.get("unit_name"),
+                    last_key.get("tool_name_id"),
+                    last_key.get("event_date"),
+                    last_key.get("event_time")
+                )
+                logger.info(
+                    f"Resuming AFTER last key: "
+                    f"({start_key_tuple[0]}, {start_key_tuple[1]}, "
+                    f"{start_key_tuple[2]}, {start_key_tuple[3]})"
+                )
+
+        if dry_run:
+            logger.info(f"[DRY RUN] Would stream partition {partition_name}")
+            return 0
+
+        # Track migration
+        state_mgr.mark_in_progress()
+        migrated_rows = 0
+        insert_buffer = []
+        buffer_size = 10000  # Larger buffer for COPY performance
+
+        # Get column order for PostgreSQL insert
+        pg_columns = self._get_pg_columns()
+
+        # Group rows by consolidation key
+        current_group: List[Dict[str, Any]] = []
+        current_key = None
+        rows_processed = 0
+
+        migration_completed = False
+        try:
+            with ProgressTracker(
+                total=None,  # Unknown total
+                description=f"Streaming {mysql_table} partition {partition_name}"
+            ) as progress:
+
+                # Log before starting to fetch (this query can take several minutes for large partitions)
+                logger.info(f"Fetching consolidation groups from MySQL partition {partition_name}...")
+
+                # Use fetch_consolidation_groups_from_partition with start_key for efficient resume
+                # MySQL will skip all keys <= start_key using WHERE clause (no unnecessary data transfer)
+                for group in mysql_conn.fetch_consolidation_groups_from_partition(
+                    mysql_table,
+                    partition_name,
+                    limit=self.settings.migration.consolidation_group_limit,
+                    offset=0,
+                    start_key=start_key_tuple  # Resume AFTER this key
+                ):
+                    rows_processed += len(group)
+
+                    # Extract consolidation key from first row
+                    if not group:
+                        continue
+
+                    first_row = group[0]
+                    key = (
+                        first_row.get("UnitName"),
+                        first_row.get("ToolNameID"),
+                        first_row.get("EventDate"),
+                        first_row.get("EventTime")
+                    )
+
+                    # No need to skip here anymore - MySQL query already handles resume
+                    # (keys are filtered in the database with WHERE clause)
+
+                    # Validate and process group
+                    self._process_group(
+                        group,
+                        key,
+                        insert_buffer,
+                        buffer_size,
+                        pg_conn,
+                        pg_table,
+                        pg_columns,
+                        state_mgr,
+                        error_logger,
+                        progress
+                    )
+
+                    # Update migrated count
+                    if len(insert_buffer) == 0:  # Just flushed
+                        migrated_rows = state_mgr.get_state().get("total_rows_migrated", migrated_rows)
+
+                # Flush remaining buffer
+                if insert_buffer:
+                    inserted = pg_conn.copy_from_with_conflict(
+                        pg_table,
+                        insert_buffer,
+                        pg_columns,
+                        conflict_columns=["unit_name", "tool_name_id", "event_timestamp", "event_year"]
+                    )
+                    migrated_rows += inserted
+                    progress.update(inserted)
+                    logger.debug(f"Final flush: {inserted} rows")
+
+            # If we reach here, migration completed successfully
+            migration_completed = True
+
+        finally:
+            # Close error logger
+            error_logger.close()
+
+            # Only mark as completed if migration actually finished
+            if migration_completed:
+                state_mgr.update_state(
+                    total_rows_migrated=migrated_rows,
+                    status="completed",
+                    mark_completed=True
+                )
+            else:
+                # Migration was interrupted - save progress but keep status as in_progress
+                logger.warning(f"Migration of partition {partition_name} was interrupted")
+                state_mgr.update_state(
+                    total_rows_migrated=migrated_rows,
+                    status="in_progress"
+                )
+
+        logger.info(
+            f"✓ Partition {partition_name} STREAMING migration complete: "
+            f"{migrated_rows} PostgreSQL rows inserted, "
+            f"{rows_processed} MySQL rows processed, "
+            f"{error_logger.get_error_count()} invalid keys skipped"
+        )
+
+        return migrated_rows
+
+    def _process_group(
+        self,
+        mysql_rows: List[Dict[str, Any]],
+        key: tuple,
+        insert_buffer: List[Dict[str, Any]],
+        buffer_size: int,
+        pg_conn: PostgreSQLConnector,
+        pg_table: str,
+        pg_columns: List[str],
+        state_mgr: StateManager,
+        error_logger: ErrorLogger,
+        progress: ProgressTracker
+    ) -> None:
+        """Process a consolidation group: validate, consolidate, and buffer.
+
+        Args:
+            mysql_rows: List of MySQL rows with same consolidation key
+            key: Consolidation key tuple (unit_name, tool_name_id, event_date, event_time)
+            insert_buffer: Buffer to add consolidated row to
+            buffer_size: Max buffer size before flush
+            pg_conn: PostgreSQL connector
+            pg_table: PostgreSQL table name
+            pg_columns: Column names for insert
+            state_mgr: State manager
+            error_logger: Error logger
+            progress: Progress tracker
+        """
+        unit_name, tool_name_id, event_date, event_time = key
+
+        # Validate consolidation key
+        is_valid, error_reason = validate_consolidation_key(
+            unit_name, tool_name_id, event_date, event_time
+        )
+
+        if not is_valid:
+            # Log invalid key and skip
+            error_logger.log_invalid_key(
+                unit_name, tool_name_id, event_date, event_time, error_reason
+            )
+            return
+
+        # Consolidate into single PostgreSQL row
+        try:
+            pg_row = consolidate_rows(self.table, mysql_rows)
+        except Exception as e:
+            logger.error(
+                f"Failed to consolidate key "
+                f"({unit_name}, {tool_name_id}, {event_date}, {event_time}): {e}"
+            )
+            error_logger.log_invalid_key(
+                unit_name, tool_name_id, event_date, event_time,
+                f"Consolidation failed: {e}"
+            )
+            return
+
+        # Add to insert buffer
+        insert_buffer.append(pg_row)
+
+        # Flush buffer when full
+        if len(insert_buffer) >= buffer_size:
+            # Use COPY with ON CONFLICT to handle UNIQUE constraint
+            # (unit_name, tool_name_id, event_timestamp, event_year)
+            inserted = pg_conn.copy_from_with_conflict(
+                pg_table,
+                insert_buffer,
+                pg_columns,
+                conflict_columns=["unit_name", "tool_name_id", "event_timestamp", "event_year"]
+            )
+            progress.update(inserted)
+
+            # Update state with last key
+            last_processed_key = {
+                "unit_name": unit_name,
+                "tool_name_id": tool_name_id,
+                "event_date": str(event_date) if event_date else None,
+                "event_time": str(event_time) if event_time else None,
+            }
+            current_total = state_mgr.get_state().get("total_rows_migrated", 0)
+            state_mgr.update_state(
+                last_key=last_processed_key,
+                total_rows_migrated=current_total + inserted
+            )
+
+            logger.debug(
+                f"Flushed {inserted} rows using COPY, "
+                f"total migrated: {current_total + inserted}"
+            )
+            insert_buffer.clear()
+
+    def migrate_partition(
+        self,
+        mysql_conn: MySQLConnector,
+        pg_conn: PostgreSQLConnector,
+        partition_name: str,
+        dry_run: bool = False,
+        resume: bool = False
+    ) -> int:
+        """Migrate a single partition from MySQL to PostgreSQL using streaming.
+
+        Args:
+            mysql_conn: MySQL connector (already connected)
+            pg_conn: PostgreSQL connector (already connected)
+            partition_name: MySQL partition name (e.g., 'p2024')
+            dry_run: If True, log what would be done without modifying data
+            resume: If True, resume from last checkpoint
+
+        Returns:
+            Number of PostgreSQL rows inserted (consolidated)
+        """
+        return self.migrate_partition_streaming(
+            mysql_conn, pg_conn, partition_name, dry_run, resume
+        )
+
+    def _get_pg_columns(self) -> list[str]:
+        """Get PostgreSQL column names in insertion order.
+
+        Returns:
+            List of column names
+        """
+        # Base columns for both tables
+        columns = [
+            "mysql_max_id",
+            "unit_name",
+            "tool_name_id",
+            "event_timestamp",
+            "event_year",
+            "measurements",
+        ]
+
+        # Add table-specific columns
+        if self.table == "RAWDATACOR":
+            columns.extend([
+                "bat_level",
+                "temperature",
+                "bat_level_module",
+                "temperature_module",
+                "rssi_module",
+                "created_at",
+            ])
+        elif self.table == "ELABDATADISP":
+            columns.extend([
+                "created_at",
+                "updated_at",
+            ])
+
+        return columns

src/migrator/state_manager.py

@@ -0,0 +1,347 @@
+"""Migration state management using PostgreSQL migration_state table.
+
+Tracks migration progress with:
+- last_key: Last consolidation key migrated (UnitName, ToolNameID, EventDate, EventTime)
+- last_completed_partition: Last partition that was fully migrated
+- total_rows_migrated: Count of PostgreSQL rows (consolidated)
+- status: pending, in_progress, completed
+"""
+from typing import Optional, Dict, Any
+from datetime import datetime, timezone
+import json
+
+from src.connectors.postgres_connector import PostgreSQLConnector
+from src.utils.logger import get_logger
+
+logger = get_logger(__name__)
+
+
+class StateManager:
+    """Manage migration state in PostgreSQL migration_state table.
+
+    Supports per-partition state tracking for parallel migration.
+    """
+
+    def __init__(self, pg_conn: PostgreSQLConnector, table_name: str, partition_name: str = "_global"):
+        """Initialize state manager.
+
+        Args:
+            pg_conn: PostgreSQL connector
+            table_name: PostgreSQL table name (rawdatacor or elabdatadisp)
+            partition_name: MySQL partition name (e.g., 'd0', 'part0') or '_global' for global state
+        """
+        self.pg_conn = pg_conn
+        self.table_name = table_name
+        self.partition_name = partition_name
+
+    def get_last_key(self) -> Optional[Dict[str, Any]]:
+        """Get last consolidation key that was migrated for this partition.
+
+        Returns:
+            Dict with keys: unit_name, tool_name_id, event_date, event_time
+            Or None if no previous migration
+        """
+        try:
+            with self.pg_conn.connection.cursor() as cursor:
+                cursor.execute(
+                    "SELECT last_key FROM migration_state WHERE table_name = %s AND partition_name = %s",
+                    (self.table_name, self.partition_name)
+                )
+                result = cursor.fetchone()
+                if result and result[0]:
+                    # last_key is stored as JSONB, psycopg returns it as dict
+                    return result[0]
+                return None
+        except Exception as e:
+            logger.debug(f"Could not get last_key from migration_state: {e}")
+            return None
+
+    def get_total_rows_migrated(self) -> int:
+        """Get total rows migrated for this partition.
+
+        Returns:
+            Count of PostgreSQL rows (consolidated)
+        """
+        try:
+            with self.pg_conn.connection.cursor() as cursor:
+                cursor.execute(
+                    "SELECT total_rows_migrated FROM migration_state WHERE table_name = %s AND partition_name = %s",
+                    (self.table_name, self.partition_name)
+                )
+                result = cursor.fetchone()
+                return result[0] if result and result[0] else 0
+        except Exception as e:
+            logger.debug(f"Could not get total_rows_migrated from migration_state: {e}")
+            return 0
+
+    def get_status(self) -> str:
+        """Get migration status for this partition.
+
+        Returns:
+            Status string: pending, in_progress, or completed
+        """
+        try:
+            with self.pg_conn.connection.cursor() as cursor:
+                cursor.execute(
+                    "SELECT status FROM migration_state WHERE table_name = %s AND partition_name = %s",
+                    (self.table_name, self.partition_name)
+                )
+                result = cursor.fetchone()
+                return result[0] if result and result[0] else "pending"
+        except Exception as e:
+            logger.debug(f"Could not get status from migration_state: {e}")
+            return "pending"
+
+    def get_state(self) -> Dict[str, Any]:
+        """Get complete migration state for this partition.
+
+        Returns:
+            Dict with keys: last_key, total_rows_migrated, status, migration_completed_at
+        """
+        try:
+            with self.pg_conn.connection.cursor() as cursor:
+                cursor.execute(
+                    "SELECT last_key, total_rows_migrated, status, migration_completed_at "
+                    "FROM migration_state WHERE table_name = %s AND partition_name = %s",
+                    (self.table_name, self.partition_name)
+                )
+                result = cursor.fetchone()
+                if result:
+                    return {
+                        "last_key": result[0],
+                        "total_rows_migrated": result[1] if result[1] else 0,
+                        "status": result[2] if result[2] else "pending",
+                        "migration_completed_at": result[3]
+                    }
+                return {
+                    "last_key": None,
+                    "total_rows_migrated": 0,
+                    "status": "pending",
+                    "migration_completed_at": None
+                }
+        except Exception as e:
+            logger.debug(f"Could not get state from migration_state: {e}")
+            return {
+                "last_key": None,
+                "total_rows_migrated": 0,
+                "status": "pending",
+                "migration_completed_at": None
+            }
+
+    def get_completed_partitions(self) -> list[str]:
+        """Get list of all completed partitions for this table.
+
+        Returns:
+            List of partition names that have been completed
+        """
+        try:
+            with self.pg_conn.connection.cursor() as cursor:
+                cursor.execute(
+                    "SELECT partition_name FROM migration_state WHERE table_name = %s AND status = 'completed'",
+                    (self.table_name,)
+                )
+                results = cursor.fetchall()
+                return [row[0] for row in results if row[0] != "_global"]
+        except Exception as e:
+            logger.debug(f"Could not get completed partitions from migration_state: {e}")
+            return []
+
+    def get_in_progress_partitions(self) -> list[str]:
+        """Get list of all in-progress partitions for this table.
+
+        Returns:
+            List of partition names that are currently in progress
+        """
+        try:
+            with self.pg_conn.connection.cursor() as cursor:
+                cursor.execute(
+                    "SELECT partition_name FROM migration_state WHERE table_name = %s AND status = 'in_progress'",
+                    (self.table_name,)
+                )
+                results = cursor.fetchall()
+                return [row[0] for row in results if row[0] != "_global"]
+        except Exception as e:
+            logger.debug(f"Could not get in-progress partitions from migration_state: {e}")
+            return []
+
+    def update_state(
+        self,
+        last_key: Optional[Dict[str, Any]] = None,
+        total_rows_migrated: Optional[int] = None,
+        status: Optional[str] = None,
+        mark_completed: bool = False
+    ) -> None:
+        """Update migration state for this partition.
+
+        Args:
+            last_key: Last consolidation key migrated
+            total_rows_migrated: Total PostgreSQL rows migrated for this partition
+            status: Migration status (pending, in_progress, completed)
+            mark_completed: If True, mark migration as completed with timestamp
+        """
+        try:
+            with self.pg_conn.connection.cursor() as cursor:
+                # Build dynamic UPDATE based on provided parameters
+                updates = []
+                params = []
+
+                if last_key is not None:
+                    updates.append("last_key = %s::jsonb")
+                    params.append(json.dumps(last_key))
+
+                if total_rows_migrated is not None:
+                    updates.append("total_rows_migrated = %s")
+                    params.append(total_rows_migrated)
+
+                if status is not None:
+                    updates.append("status = %s")
+                    params.append(status)
+
+                if mark_completed:
+                    updates.append("migration_completed_at = %s")
+                    params.append(datetime.now(timezone.utc))
+                    if status is None:
+                        updates.append("status = 'completed'")
+
+                if not updates:
+                    logger.warning("update_state called with no parameters to update")
+                    return
+
+                # Upsert: INSERT or UPDATE
+                query = f"""
+                    INSERT INTO migration_state (table_name, partition_name, {', '.join([u.split('=')[0].strip() for u in updates])})
+                    VALUES (%s, %s, {', '.join(['%s'] * len(updates))})
+                    ON CONFLICT (table_name, partition_name) DO UPDATE SET
+                        {', '.join(updates)}
+                """
+
+                all_params = [self.table_name, self.partition_name] + params + params
+                cursor.execute(query, tuple(all_params))
+                self.pg_conn.connection.commit()
+
+                logger.debug(
+                    f"Migration state updated for {self.table_name}/{self.partition_name}: "
+                    f"last_key={last_key}, total={total_rows_migrated}, status={status}"
+                )
+
+        except Exception as e:
+            logger.error(f"Failed to update migration state: {e}")
+            self.pg_conn.connection.rollback()
+            raise
+
+    def reset_state(self) -> None:
+        """Reset migration state for this partition.
+
+        Deletes the state record for this specific partition.
+        """
+        try:
+            with self.pg_conn.connection.cursor() as cursor:
+                cursor.execute(
+                    "DELETE FROM migration_state WHERE table_name = %s AND partition_name = %s",
+                    (self.table_name, self.partition_name)
+                )
+                self.pg_conn.connection.commit()
+                logger.info(f"Migration state reset for {self.table_name}/{self.partition_name}")
+        except Exception as e:
+            logger.error(f"Failed to reset migration state: {e}")
+            self.pg_conn.connection.rollback()
+            raise
+
+    def mark_in_progress(self) -> None:
+        """Mark migration as in progress."""
+        self.update_state(status="in_progress")
+
+    def mark_completed(self, total_rows: int) -> None:
+        """Mark migration as completed.
+
+        Args:
+            total_rows: Final count of migrated rows
+        """
+        self.update_state(
+            total_rows_migrated=total_rows,
+            status="completed",
+            mark_completed=True
+        )
+
+    def get_max_last_key_across_partitions(self) -> Optional[Dict[str, Any]]:
+        """Get the maximum (latest) last_key across all partitions for this table.
+
+        Used when completing full migration to set the global last_key for incremental migration.
+
+        The consolidation key ordering is: (UnitName, ToolNameID, EventDate, EventTime)
+        To find the maximum, we order by these fields in DESC order.
+
+        Returns:
+            The latest last_key across all partitions, or None if no partitions have keys
+        """
+        try:
+            with self.pg_conn.connection.cursor() as cursor:
+                # Order by the full consolidation key (UnitName, ToolNameID, EventDate, EventTime) DESC
+                # This matches the ordering used in MySQL fetch_consolidation_keys_after
+                cursor.execute(
+                    """
+                    SELECT last_key
+                    FROM migration_state
+                    WHERE table_name = %s AND partition_name != '_global' AND last_key IS NOT NULL
+                    ORDER BY
+                        last_key->>'unit_name' DESC,
+                        last_key->>'tool_name_id' DESC,
+                        (last_key->>'event_date')::date DESC,
+                        (last_key->>'event_time')::time DESC
+                    LIMIT 1
+                    """,
+                    (self.table_name,)
+                )
+                result = cursor.fetchone()
+                if result and result[0]:
+                    return result[0]
+                return None
+        except Exception as e:
+            logger.debug(f"Could not get max last_key across partitions: {e}")
+            return None
+
+    def get_max_last_key_from_data(self) -> Optional[Dict[str, Any]]:
+        """Get the consolidation key of the most recent event in PostgreSQL.
+
+        This queries the target table directly (not migration_state) to find the row with
+        the maximum event_timestamp (most recent event), then returns its consolidation key.
+
+        We order by event_timestamp only (not the full consolidation key) because:
+        1. Ordering by unit_name can pick up corrupted data (Java error strings)
+        2. event_timestamp represents the actual chronological order of events
+        3. This avoids re-processing old data that happens to sort later alphabetically
+
+        Corrupted data (like '[Ljava.lang.String;@...' in unit_name) is explicitly excluded.
+
+        Returns:
+            The consolidation key of the most recent event, or None if table is empty
+        """
+        try:
+            with self.pg_conn.connection.cursor() as cursor:
+                # Find the row with maximum event_timestamp (most recent event)
+                # We use ONLY event_timestamp because ordering by unit_name can pick up
+                # corrupted data (like Java error strings) which sort incorrectly
+                cursor.execute(
+                    f"""
+                    SELECT unit_name, tool_name_id,
+                           DATE(event_timestamp)::text as event_date,
+                           event_timestamp::time::text as event_time,
+                           event_timestamp
+                    FROM {self.table_name}
+                    WHERE unit_name NOT LIKE '[L%'  -- Exclude corrupted Java strings
+                    ORDER BY event_timestamp DESC
+                    LIMIT 1
+                    """
+                )
+                result = cursor.fetchone()
+                if result:
+                    return {
+                        "unit_name": result[0],
+                        "tool_name_id": result[1],
+                        "event_date": result[2],
+                        "event_time": result[3]
+                    }
+                return None
+        except Exception as e:
+            logger.error(f"Could not get max last_key from data table {self.table_name}: {e}")
+            return None

src/transformers/schema_transformer.py

@@ -1,4 +1,12 @@
-"""PostgreSQL schema creation from MySQL structure."""
+"""PostgreSQL schema creation for MySQL to PostgreSQL migration.
+
+New design:
+- id: BIGSERIAL PRIMARY KEY (auto-increment)
+- mysql_max_id: max(idElabData) from consolidated MySQL records
+- event_timestamp: TIMESTAMP created from MySQL EventDate + EventTime
+- Consolidation key MySQL: (UnitName, ToolNameID, EventDate, EventTime)
+- NodeNum consolidated in measurements JSONB
+"""
 from config import PARTITION_YEARS
 from src.utils.logger import get_logger
 
@@ -8,31 +16,37 @@ logger = get_logger(__name__)
 def create_rawdatacor_schema() -> str:
     """Create PostgreSQL schema for RAWDATACOR table.
 
+    Schema design:
+    - id: Auto-increment primary key (PostgreSQL sequence)
+    - mysql_max_id: Max ID from MySQL records that were consolidated
+    - Consolidation: Multiple MySQL rows (different NodeNum) → 1 PostgreSQL row
+    - measurements JSONB structure:
+      {
+        "node_1": {"0": {"value": "123.45", "unit": "°C"}, ...},
+        "node_2": {"0": {"value": "67.89", "unit": "bar"}, ...},
+        ...
+      }
+
     Returns:
         SQL script to create the table with partitions
     """
     sql = """
--- Create sequence for id auto-increment
-CREATE SEQUENCE IF NOT EXISTS rawdatacor_id_seq;
-
--- Create RAWDATACOR table with partitioning
--- Note: node_num is stored in measurements JSONB, not as a separate column
+-- Create RAWDATACOR table with partitioning by year
 CREATE TABLE IF NOT EXISTS rawdatacor (
-    id BIGINT NOT NULL DEFAULT nextval('rawdatacor_id_seq'),
-    unit_name VARCHAR(32),
-    tool_name_id VARCHAR(32) NOT NULL,
+    id bigint GENERATED BY DEFAULT AS IDENTITY,
+    mysql_max_id INTEGER,
+    unit_name VARCHAR(50) NOT NULL,
+    tool_name_id VARCHAR(50) NOT NULL,
     event_timestamp TIMESTAMP NOT NULL,
-    bat_level NUMERIC(4,2) NOT NULL,
-    temperature NUMERIC(5,2) NOT NULL,
-    measurements JSONB,
-    created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
+    event_year smallint NOT NULL,
+    measurements JSONB NOT NULL,
+    bat_level NUMERIC(4,2),
+    temperature NUMERIC(5,2),
     bat_level_module NUMERIC(4,2),
     temperature_module NUMERIC(5,2),
-    rssi_module INTEGER
-) PARTITION BY RANGE (EXTRACT(YEAR FROM event_timestamp));
-
--- Note: PostgreSQL doesn't support PRIMARY KEY or UNIQUE constraints
--- with RANGE partitioning on expressions. Using sequence for id auto-increment.
+    rssi_module INTEGER,
+    created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP
+) PARTITION BY RANGE (event_year);
 
 -- Create partitions for each year
 """
@@ -54,21 +68,21 @@ CREATE TABLE IF NOT EXISTS rawdatacor_default
 
     # Add indexes
     sql += """
--- Create indexes
-CREATE INDEX IF NOT EXISTS idx_unit_tool_datetime_raw
-    ON rawdatacor(unit_name, tool_name_id, event_timestamp);
+-- Create indexes for efficient queries
+-- UNIQUE constraint on (id, event_year) for partitioned table primary key
+CREATE UNIQUE INDEX IF NOT EXISTS rawdatacor_pkey
+    ON rawdatacor(id, event_year);
 
-CREATE INDEX IF NOT EXISTS idx_unit_tool_raw
+-- UNIQUE constraint on consolidation key to prevent duplicates
+-- This is the key used to consolidate MySQL rows (UnitName, ToolNameID, EventDate, EventTime)
+CREATE UNIQUE INDEX IF NOT EXISTS rawdatacor_consolidation_key_unique
+    ON rawdatacor(unit_name, tool_name_id, event_timestamp, event_year);
+
+CREATE INDEX IF NOT EXISTS idx_rawdatacor_unit_tool
     ON rawdatacor(unit_name, tool_name_id);
 
-CREATE INDEX IF NOT EXISTS idx_measurements_gin_raw
+CREATE INDEX IF NOT EXISTS idx_rawdatacor_measurements_gin
     ON rawdatacor USING GIN (measurements);
-
-CREATE INDEX IF NOT EXISTS idx_event_timestamp_raw
-    ON rawdatacor(event_timestamp);
-
-CREATE INDEX IF NOT EXISTS idx_id_raw
-    ON rawdatacor(id);
 """
 
     return sql
@@ -77,27 +91,39 @@ CREATE INDEX IF NOT EXISTS idx_id_raw
 def create_elabdatadisp_schema() -> str:
     """Create PostgreSQL schema for ELABDATADISP table.
 
+    Schema design:
+    - id: Auto-increment primary key (PostgreSQL sequence)
+    - mysql_max_id: Max idElabData from MySQL records that were consolidated
+    - Consolidation: Multiple MySQL rows (different NodeNum) → 1 PostgreSQL row
+    - measurements JSONB structure:
+      {
+        "node_1": {
+          "shifts": {"x": 1.234, "y": 2.345, ...},
+          "coordinates": {"x": 10.123, "y": 20.234, ...},
+          "kinematics": {...},
+          "sensors": {...},
+          "calculated": {...}
+        },
+        "node_2": { ... },
+        ...
+      }
+
     Returns:
         SQL script to create the table with partitions
     """
     sql = """
--- Create sequence for id_elab_data auto-increment
-CREATE SEQUENCE IF NOT EXISTS elabdatadisp_id_seq;
-
--- Create ELABDATADISP table with partitioning
--- Note: node_num, state, and calc_err are stored in measurements JSONB, not as separate columns
+-- Create ELABDATADISP table with partitioning by year
 CREATE TABLE IF NOT EXISTS elabdatadisp (
-    id_elab_data BIGINT NOT NULL DEFAULT nextval('elabdatadisp_id_seq'),
-    unit_name VARCHAR(32),
-    tool_name_id VARCHAR(32) NOT NULL,
+    id bigint GENERATED BY DEFAULT AS IDENTITY,
+    mysql_max_id INTEGER NOT NULL,
+    unit_name VARCHAR(50),
+    tool_name_id VARCHAR(50),
     event_timestamp TIMESTAMP NOT NULL,
-    measurements JSONB,
+    event_year smallint NOT NULL,
+    measurements JSONB NOT NULL,
     created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
     updated_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP
-) PARTITION BY RANGE (EXTRACT(YEAR FROM event_timestamp));
-
--- Note: PostgreSQL doesn't support PRIMARY KEY or UNIQUE constraints
--- with RANGE partitioning on expressions. Using sequence for id_elab_data auto-increment.
+) PARTITION BY RANGE (event_year);
 
 -- Create partitions for each year
 """
@@ -119,21 +145,21 @@ CREATE TABLE IF NOT EXISTS elabdatadisp_default
 
     # Add indexes
     sql += """
--- Create indexes
-CREATE INDEX IF NOT EXISTS idx_unit_tool_datetime_elab
-    ON elabdatadisp(unit_name, tool_name_id, event_timestamp);
+-- Create indexes for efficient queries
+-- UNIQUE constraint on (id, event_year) for partitioned table primary key
+CREATE UNIQUE INDEX IF NOT EXISTS elabdatadisp_pkey
+    ON elabdatadisp(id, event_year);
 
-CREATE INDEX IF NOT EXISTS idx_unit_tool_elab
+-- UNIQUE constraint on consolidation key to prevent duplicates
+-- This is the key used to consolidate MySQL rows (UnitName, ToolNameID, EventDate, EventTime)
+CREATE UNIQUE INDEX IF NOT EXISTS elabdatadisp_consolidation_key_unique
+    ON elabdatadisp(unit_name, tool_name_id, event_timestamp, event_year);
+
+CREATE INDEX IF NOT EXISTS idx_elabdatadisp_unit_tool
     ON elabdatadisp(unit_name, tool_name_id);
 
-CREATE INDEX IF NOT EXISTS idx_measurements_gin_elab
+CREATE INDEX IF NOT EXISTS idx_elabdatadisp_measurements_gin
     ON elabdatadisp USING GIN (measurements);
-
-CREATE INDEX IF NOT EXISTS idx_event_timestamp_elab
-    ON elabdatadisp(event_timestamp);
-
-CREATE INDEX IF NOT EXISTS idx_id_elab_data
-    ON elabdatadisp(id_elab_data);
 """
 
     return sql
@@ -142,21 +168,42 @@ CREATE INDEX IF NOT EXISTS idx_id_elab_data
 def create_migration_state_table() -> str:
     """Create table to track migration state.
 
+    Tracks migration progress per partition for parallel processing:
+    - table_name + partition_name: Composite primary key for per-partition tracking
+    - last_key: Last consolidated key migrated (unit_name, tool_name_id, event_date, event_time)
+    - total_rows_migrated: Count of PostgreSQL rows (consolidated) for this partition
+    - status: pending, in_progress, completed
+
     Returns:
         SQL to create migration_state table
     """
     sql = """
--- Create table to track migration state
+-- Create table to track migration state (supports parallel partition migration)
 CREATE TABLE IF NOT EXISTS migration_state (
-    table_name VARCHAR(255) PRIMARY KEY,
-    last_migrated_timestamp TIMESTAMP,
-    last_migrated_id BIGINT,
-    migration_started_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
-    migration_completed_at TIMESTAMP,
+    table_name VARCHAR(255) NOT NULL,
+    partition_name VARCHAR(255) NOT NULL,
+    last_key JSONB,
+    migration_started_at TIMESTAMPTZ DEFAULT CURRENT_TIMESTAMP,
+    migration_completed_at TIMESTAMPTZ,
     total_rows_migrated BIGINT DEFAULT 0,
     status VARCHAR(32) DEFAULT 'pending',
-    last_completed_partition VARCHAR(255)
+    PRIMARY KEY (table_name, partition_name),
+    CONSTRAINT last_key_structure CHECK (
+        last_key IS NULL OR (
+            last_key ? 'unit_name' AND
+            last_key ? 'tool_name_id' AND
+            last_key ? 'event_date' AND
+            last_key ? 'event_time'
+        )
+    )
 );
+
+-- Index for querying migration state
+CREATE INDEX IF NOT EXISTS idx_migration_state_status
+    ON migration_state(table_name, status);
+
+CREATE INDEX IF NOT EXISTS idx_migration_state_partition
+    ON migration_state(partition_name);
 """
     return sql
 

src/utils/progress.py

@@ -1,73 +1,109 @@
-"""Progress tracking utility."""
-from rich.progress import (
-    Progress,
-    SpinnerColumn,
-    BarColumn,
-    TaskProgressColumn,
-    TimeRemainingColumn,
-    TimeElapsedColumn,
-    TransferSpeedColumn,
-)
-from rich.console import Console
+"""Progress tracking utility - lightweight logging version."""
 import time
+from typing import Optional
+from src.utils.logger import get_logger
+
+logger = get_logger(__name__)
 
 
 class ProgressTracker:
-    """Track migration progress with Rich progress bar."""
+    """Track migration progress with periodic text logging (lightweight)."""
 
-    def __init__(self, total: int, description: str = "Migrating"):
+    def __init__(self, total: Optional[int] = None, description: str = "Migrating", log_interval: Optional[int] = None):
         """Initialize progress tracker.
 
         Args:
-            total: Total number of items to process
+            total: Total number of items to process (None if unknown)
             description: Description of the task
+            log_interval: Log progress every N items (if None, reads from config)
         """
         self.total = total
         self.description = description
-        self.progress = Progress(
-            SpinnerColumn(),
-            BarColumn(),
-            TaskProgressColumn(),
-            TimeElapsedColumn(),
-            TimeRemainingColumn(),
-            TransferSpeedColumn(),
-            console=Console(),
-        )
-        self.task_id = None
+
+        # Read log_interval from config if not provided
+        if log_interval is None:
+            from config import get_settings
+            self.log_interval = get_settings().migration.progress_log_interval
+        else:
+            self.log_interval = log_interval
+
         self.start_time = None
         self.processed = 0
+        self.last_log_count = 0
 
     def __enter__(self):
         """Context manager entry."""
-        self.progress.start()
-        self.task_id = self.progress.add_task(
-            self.description, total=self.total
-        )
         self.start_time = time.time()
+        logger.info(f"Starting: {self.description}")
         return self
 
     def __exit__(self, exc_type, exc_val, exc_tb):
         """Context manager exit."""
-        self.progress.stop()
         if exc_type is None:
             elapsed = time.time() - self.start_time
             rate = self.processed / elapsed if elapsed > 0 else 0
-            self.progress.console.print(
-                f"[green]✓ Completed: {self.processed}/{self.total} items "
-                f"in {elapsed:.2f}s ({rate:.0f} items/sec)[/green]"
-            )
+
+            # Format elapsed time
+            if elapsed < 60:
+                elapsed_str = f"{elapsed:.1f}s"
+            elif elapsed < 3600:
+                elapsed_str = f"{elapsed/60:.1f}m"
+            else:
+                elapsed_str = f"{elapsed/3600:.1f}h"
+
+            if self.total is not None:
+                logger.info(
+                    f"✓ Completed: {self.processed:,}/{self.total:,} items "
+                    f"in {elapsed_str} ({rate:.0f} items/sec)"
+                )
+            else:
+                logger.info(
+                    f"✓ Completed: {self.processed:,} items "
+                    f"in {elapsed_str} ({rate:.0f} items/sec)"
+                )
 
     def update(self, advance: int = 1):
-        """Update progress.
+        """Update progress and log periodically.
 
         Args:
             advance: Number of items processed
         """
-        if self.task_id is not None:
-            self.progress.update(self.task_id, advance=advance)
-            self.processed += advance
+        self.processed += advance
+
+        # Log every log_interval items
+        if self.processed - self.last_log_count >= self.log_interval:
+            self._log_progress()
+            self.last_log_count = self.processed
+
+    def _log_progress(self):
+        """Log current progress."""
+        elapsed = time.time() - self.start_time
+        rate = self.processed / elapsed if elapsed > 0 else 0
+
+        # Format elapsed time
+        if elapsed < 60:
+            elapsed_str = f"{elapsed:.0f}s"
+        elif elapsed < 3600:
+            mins = int(elapsed / 60)
+            secs = int(elapsed % 60)
+            elapsed_str = f"{mins}m {secs}s"
+        else:
+            hours = int(elapsed / 3600)
+            mins = int((elapsed % 3600) / 60)
+            elapsed_str = f"{hours}h {mins}m"
+
+        if self.total is not None:
+            progress_pct = (self.processed / self.total) * 100
+            logger.info(
+                f"{self.description}: {self.processed:,}/{self.total:,} items "
+                f"({progress_pct:.1f}%) - elapsed: {elapsed_str} - rate: {rate:.0f} items/sec"
+            )
+        else:
+            logger.info(
+                f"{self.description}: {self.processed:,} items "
+                f"- elapsed: {elapsed_str} - rate: {rate:.0f} items/sec"
+            )
 
     def print_status(self, message: str):
-        """Print a status message without interrupting progress bar."""
-        if self.task_id is not None:
-            self.progress.print(message)
+        """Print a status message."""
+        logger.info(message)

src/utils/validation.py

@@ -0,0 +1,170 @@
+"""Data validation utilities for migration."""
+from typing import Dict, Any, Optional, Tuple
+from datetime import datetime, date
+import os
+from src.utils.logger import get_logger
+
+logger = get_logger(__name__)
+
+
+class ErrorLogger:
+    """Log invalid migration keys to a file."""
+
+    def __init__(self, table: str, partition: str, use_timestamp: bool = False):
+        """Initialize error logger.
+
+        Args:
+            table: Table name
+            partition: Partition name (e.g., 'p2024' or 'incremental')
+            use_timestamp: If True, add timestamp to filename (for incremental migrations)
+        """
+        self.table = table
+        self.partition = partition
+
+        # Add timestamp to filename for incremental migrations to avoid overwriting
+        if use_timestamp or partition == "incremental":
+            timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
+            self.error_file = f"migration_errors_{table}_{partition}_{timestamp}.log"
+        else:
+            self.error_file = f"migration_errors_{table}_{partition}.log"
+
+        self.error_count = 0
+
+        # Create error file with header
+        with open(self.error_file, "w") as f:
+            f.write(f"# Migration errors for {table} partition {partition}\n")
+            f.write(f"# Timestamp: {datetime.now().strftime('%Y-%m-%d %H:%M:%S')}\n")
+            f.write("# Format: UnitName|ToolNameID|EventDate|EventTime|Reason\n\n")
+
+        logger.info(f"Error log file created: {self.error_file}")
+
+    def log_invalid_key(
+        self,
+        unit_name: Any,
+        tool_name_id: Any,
+        event_date: Any,
+        event_time: Any,
+        reason: str
+    ) -> None:
+        """Log an invalid consolidation key.
+
+        Args:
+            unit_name: UnitName value
+            tool_name_id: ToolNameID value
+            event_date: EventDate value
+            event_time: EventTime value
+            reason: Reason for rejection
+        """
+        with open(self.error_file, "a") as f:
+            f.write(f"{unit_name}|{tool_name_id}|{event_date}|{event_time}|{reason}\n")
+
+        self.error_count += 1
+
+        if self.error_count % 100 == 0:
+            logger.warning(f"Logged {self.error_count} invalid keys to {self.error_file}")
+
+    def get_error_count(self) -> int:
+        """Get total number of errors logged.
+
+        Returns:
+            Number of errors logged
+        """
+        return self.error_count
+
+    def close(self) -> None:
+        """Close error logger and log summary."""
+        if self.error_count > 0:
+            logger.warning(
+                f"Total invalid keys for {self.table} partition {self.partition}: "
+                f"{self.error_count} (see {self.error_file})"
+            )
+        else:
+            logger.info(f"No invalid keys found for {self.table} partition {self.partition}")
+            # Remove empty error file
+            if os.path.exists(self.error_file):
+                os.remove(self.error_file)
+
+
+def validate_consolidation_key(
+    unit_name: Any,
+    tool_name_id: Any,
+    event_date: Any,
+    event_time: Any
+) -> Tuple[bool, Optional[str]]:
+    """Validate a consolidation key.
+
+    Args:
+        unit_name: UnitName value
+        tool_name_id: ToolNameID value
+        event_date: EventDate value
+        event_time: EventTime value
+
+    Returns:
+        Tuple of (is_valid, error_reason)
+        If valid: (True, None)
+        If invalid: (False, "reason description")
+    """
+    # Check for NULL unit_name or tool_name_id
+    if unit_name is None or unit_name == "":
+        return False, "UnitName is NULL or empty"
+
+    # Check for corrupted Java strings (like '[Ljava.lang.String;@...')
+    if isinstance(unit_name, str) and unit_name.startswith("[L"):
+        return False, f"UnitName is corrupted Java string: {unit_name}"
+
+    if tool_name_id is None or tool_name_id == "":
+        return False, "ToolNameID is NULL or empty"
+
+    # Check for NULL or invalid dates
+    if event_date is None:
+        return False, "EventDate is NULL"
+
+    # Check for invalid date like '0000-00-00'
+    try:
+        if isinstance(event_date, str):
+            if event_date.startswith("0000-00-00"):
+                return False, f"EventDate is invalid: {event_date}"
+            # Try to parse
+            parsed_date = datetime.strptime(event_date, "%Y-%m-%d").date()
+        elif isinstance(event_date, (date, datetime)):
+            parsed_date = event_date if isinstance(event_date, date) else event_date.date()
+            # Check for zero date
+            if parsed_date.year == 0:
+                return False, f"EventDate year is 0: {event_date}"
+        else:
+            return False, f"EventDate has invalid type: {type(event_date)}"
+    except (ValueError, AttributeError) as e:
+        return False, f"EventDate parsing failed: {event_date} ({e})"
+
+    # Check for NULL event_time
+    if event_time is None:
+        return False, "EventTime is NULL"
+
+    return True, None
+
+
+def validate_mysql_row(row: Dict[str, Any]) -> Tuple[bool, Optional[str]]:
+    """Validate a complete MySQL row for migration.
+
+    Args:
+        row: MySQL row dictionary
+
+    Returns:
+        Tuple of (is_valid, error_reason)
+    """
+    # Validate consolidation key
+    is_valid, reason = validate_consolidation_key(
+        row.get("UnitName"),
+        row.get("ToolNameID"),
+        row.get("EventDate"),
+        row.get("EventTime")
+    )
+
+    if not is_valid:
+        return False, reason
+
+    # Check for NodeNum
+    if row.get("NodeNum") is None:
+        return False, "NodeNum is NULL"
+
+    return True, None

test_generator_output.py

@@ -0,0 +1,63 @@
+#!/usr/bin/env python3
+"""Debug what the generator is actually returning."""
+import sys
+sys.path.insert(0, '/home/alex/devel/mysql2postgres')
+
+from src.connectors.mysql_connector import MySQLConnector
+from src.utils.logger import setup_logger, get_logger
+
+setup_logger(__name__)
+logger = get_logger(__name__)
+
+print("\n" + "="*80)
+print("Testing consolidation groups generator for d1")
+print("="*80 + "\n")
+
+with MySQLConnector() as mysql_conn:
+    partition = "d1"
+    group_num = 0
+    # Use datetime objects to match what the generator uses
+    import datetime
+    target_key = ("ID0003", "DT0002", datetime.date(2014, 8, 31), datetime.timedelta(hours=11, minutes=59, seconds=10))
+
+    print("First 20 groups from generator:\n")
+    print("DEBUG: First row columns:", flush=True)
+
+    for group_rows in mysql_conn.fetch_consolidation_groups_from_partition(
+        "ELABDATADISP",
+        partition,
+        limit=100
+    ):
+        group_num += 1
+        if group_rows:
+            first_row = group_rows[0]
+
+            # Debug: print all columns from first group
+            if group_num == 1:
+                print(f"  Available columns: {first_row.keys()}\n")
+                print(f"  First row data: {dict(first_row)}\n")
+
+            key = (
+                first_row.get("UnitName"),
+                first_row.get("ToolNameID"),
+                str(first_row.get("EventDate")),
+                str(first_row.get("EventTime"))
+            )
+            nodes = sorted([r.get('NodeNum') for r in group_rows])
+
+            # Show first 20 groups or target key
+            if group_num <= 20 or key == target_key:
+                print(f"Group {group_num}: key={key}")
+                print(f"  Nodes ({len(nodes)}): {nodes}")
+                print(f"  Rows count: {len(group_rows)}\n")
+
+            if key == target_key:
+                print("^^^ THIS IS THE TARGET KEY! ^^^\n")
+                break
+
+        if group_num >= 100:
+            print(f"\nStopped at group {group_num}")
+            break
+
+print(f"\nTotal groups processed: {group_num}")
+print("Done!\n")

test_target_record.py

@@ -0,0 +1,90 @@
+#!/usr/bin/env python3
+"""Test if target record is being consolidated correctly."""
+from src.connectors.mysql_connector import MySQLConnector
+from src.transformers.data_transformer import DataTransformer
+from src.utils.logger import setup_logger, get_logger
+
+setup_logger(__name__)
+logger = get_logger(__name__)
+
+print("\n" + "="*80)
+print("Testing target record consolidation")
+print("="*80 + "\n")
+
+target_key = ("M1_ID0246", "DT0001", "2023-06-26", "10:43:59")
+
+with MySQLConnector() as mysql_conn:
+    partition = "d10"
+    group_num = 0
+    found = False
+
+    print("Fetching consolidation groups from d10...\n")
+
+    for group_rows in mysql_conn.fetch_consolidation_groups_from_partition(
+        "ELABDATADISP",
+        partition,
+        limit=100
+    ):
+        group_num += 1
+        if group_rows:
+            first_row = group_rows[0]
+            key = (
+                first_row.get("UnitName"),
+                first_row.get("ToolNameID"),
+                str(first_row.get("EventDate")),
+                str(first_row.get("EventTime"))
+            )
+            nodes = sorted([r.get('NodeNum') for r in group_rows])
+
+            # Show first 10 groups
+            if group_num <= 10:
+                print(f"Group {group_num}: key={key}, nodes={len(nodes)} items")
+
+            if key == target_key:
+                print(f"\n✓ FOUND TARGET KEY in group {group_num}!")
+                print(f"  Key: {key}")
+                print(f"  Nodes: {nodes}")
+                print(f"  Count: {len(group_rows)}")
+
+                if len(nodes) == 22 and nodes == list(range(1, 23)):
+                    print("\n✓ All 22 nodes present!")
+
+                    # Test consolidation
+                    consolidated = DataTransformer.consolidate_elabdatadisp_batch(group_rows)
+                    print(f"Consolidated to {len(consolidated)} row(s)")
+
+                    if len(consolidated) == 1:
+                        print("✓ Consolidated to 1 row!")
+                        import json
+                        meas = consolidated[0].get("measurements")
+                        if isinstance(meas, str):
+                            meas = json.loads(meas)
+                        cons_nodes = sorted([int(k) for k in meas.keys()])
+                        print(f"Measurements nodes: {cons_nodes}")
+
+                        if cons_nodes == list(range(1, 23)):
+                            print("\n" + "="*80)
+                            print("✓✓✓ TARGET RECORD CONSOLIDATES CORRECTLY ✓✓✓")
+                            print("="*80)
+                        else:
+                            print(f"✗ Expected nodes 1-22, got {cons_nodes}")
+                    else:
+                        print(f"✗ Expected 1 consolidated row, got {len(consolidated)}")
+                else:
+                    print(f"✗ INCOMPLETE! Expected 22 nodes, got {len(nodes)}")
+                    print(f"  Expected: {list(range(1, 23))}")
+                    print(f"  Got: {nodes}")
+
+                found = True
+                break
+
+        # Safety limit
+        if group_num >= 1000:
+            print(f"\nStopped at group {group_num} (safety limit)")
+            break
+
+if not found:
+    print(f"\n✗ Target key NOT FOUND in first {group_num} groups")
+    print("\nThis is a PROBLEM - the record is not being returned by the generator!")
+
+print("\nDone!\n")

web_ui.py

@@ -0,0 +1,593 @@
+"""Web UI for MySQL to PostgreSQL migration monitoring and control.
+
+Provides a Gradio-based interface for:
+- Viewing migration status and progress
+- Starting/monitoring migrations
+- Viewing logs and errors
+- Performance metrics and graphs
+"""
+import gradio as gr
+import pandas as pd
+import plotly.graph_objects as go
+import plotly.express as px
+from pathlib import Path
+import threading
+import time
+from datetime import datetime
+from typing import Optional, Dict, List, Tuple
+import logging
+
+from config import get_settings
+from src.connectors.postgres_connector import PostgreSQLConnector
+from src.migrator.full_migrator import run_full_migration
+from src.migrator.parallel_migrator import run_parallel_migration
+from src.utils.logger import get_logger
+
+logger = get_logger(__name__)
+
+# Global state for tracking running migrations
+running_migrations = {}
+migration_logs = {}
+
+
+def get_migration_state_df() -> pd.DataFrame:
+    """Fetch current migration state from PostgreSQL.
+
+    Returns:
+        DataFrame with columns: table_name, partition_name, status,
+        total_rows_migrated, migration_started_at, migration_completed_at
+    """
+    try:
+        with PostgreSQLConnector() as pg_conn:
+            with pg_conn.connection.cursor() as cursor:
+                cursor.execute("""
+                    SELECT
+                        table_name,
+                        partition_name,
+                        status,
+                        total_rows_migrated,
+                        migration_started_at,
+                        migration_completed_at,
+                        last_key
+                    FROM migration_state
+                    WHERE partition_name != '_global'
+                    ORDER BY table_name, partition_name
+                """)
+
+                rows = cursor.fetchall()
+
+                if not rows:
+                    return pd.DataFrame(columns=[
+                        'Table', 'Partition', 'Status', 'Rows Migrated',
+                        'Started At', 'Completed At'
+                    ])
+
+                df = pd.DataFrame(rows, columns=[
+                    'Table', 'Partition', 'Status', 'Rows Migrated',
+                    'Started At', 'Completed At', 'Last Key'
+                ])
+
+                # Format dates
+                df['Started At'] = pd.to_datetime(df['Started At']).dt.strftime('%Y-%m-%d %H:%M:%S')
+                df['Completed At'] = pd.to_datetime(df['Completed At']).dt.strftime('%Y-%m-%d %H:%M:%S')
+                df['Completed At'] = df['Completed At'].fillna('-')
+
+                # Drop Last Key column for display (too verbose)
+                df = df.drop('Last Key', axis=1)
+
+                return df
+
+    except Exception as e:
+        logger.error(f"Failed to fetch migration state: {e}")
+        return pd.DataFrame(columns=[
+            'Table', 'Partition', 'Status', 'Rows Migrated',
+            'Started At', 'Completed At'
+        ])
+
+
+def get_migration_summary() -> Dict[str, any]:
+    """Get summary statistics for migrations.
+
+    Returns:
+        Dict with total/completed/in_progress/pending counts per table
+    """
+    try:
+        with PostgreSQLConnector() as pg_conn:
+            with pg_conn.connection.cursor() as cursor:
+                cursor.execute("""
+                    SELECT
+                        table_name,
+                        status,
+                        COUNT(*) as count,
+                        SUM(total_rows_migrated) as total_rows
+                    FROM migration_state
+                    WHERE partition_name != '_global'
+                    GROUP BY table_name, status
+                    ORDER BY table_name, status
+                """)
+
+                rows = cursor.fetchall()
+
+                summary = {
+                    'RAWDATACOR': {'completed': 0, 'in_progress': 0, 'pending': 0, 'total_rows': 0},
+                    'ELABDATADISP': {'completed': 0, 'in_progress': 0, 'pending': 0, 'total_rows': 0}
+                }
+
+                for table, status, count, total_rows in rows:
+                    table_upper = table.upper()
+                    if table_upper in summary:
+                        summary[table_upper][status] = count
+                        summary[table_upper]['total_rows'] += total_rows or 0
+
+                return summary
+
+    except Exception as e:
+        logger.error(f"Failed to fetch migration summary: {e}")
+        return {
+            'RAWDATACOR': {'completed': 0, 'in_progress': 0, 'pending': 0, 'total_rows': 0},
+            'ELABDATADISP': {'completed': 0, 'in_progress': 0, 'pending': 0, 'total_rows': 0}
+        }
+
+
+def create_status_chart() -> go.Figure:
+    """Create a bar chart showing migration status by table.
+
+    Returns:
+        Plotly figure
+    """
+    summary = get_migration_summary()
+
+    tables = []
+    completed = []
+    in_progress = []
+    pending = []
+
+    for table, stats in summary.items():
+        tables.append(table)
+        completed.append(stats['completed'])
+        in_progress.append(stats['in_progress'])
+        pending.append(stats['pending'])
+
+    fig = go.Figure(data=[
+        go.Bar(name='Completed', x=tables, y=completed, marker_color='green'),
+        go.Bar(name='In Progress', x=tables, y=in_progress, marker_color='orange'),
+        go.Bar(name='Pending', x=tables, y=pending, marker_color='gray')
+    ])
+
+    fig.update_layout(
+        barmode='stack',
+        title='Migration Status by Table',
+        xaxis_title='Table',
+        yaxis_title='Number of Partitions',
+        height=400
+    )
+
+    return fig
+
+
+def get_error_logs() -> List[Tuple[str, str]]:
+    """Get list of error log files with their paths.
+
+    Returns:
+        List of (filename, filepath) tuples
+    """
+    error_logs = []
+    for log_file in Path('.').glob('migration_errors_*.log'):
+        error_logs.append((log_file.name, str(log_file)))
+
+    return sorted(error_logs, key=lambda x: x[0], reverse=True)
+
+
+def read_error_log(log_file: str) -> str:
+    """Read contents of an error log file.
+
+    Args:
+        log_file: Path to log file
+
+    Returns:
+        Contents of log file
+    """
+    try:
+        if not log_file or log_file == "Select a log file...":
+            return "No log file selected"
+
+        with open(log_file, 'r') as f:
+            content = f.read()
+
+        if not content:
+            return f"Log file {log_file} is empty (no errors logged)"
+
+        return content
+
+    except Exception as e:
+        return f"Error reading log file: {e}"
+
+
+def start_migration_task(
+    table: str,
+    parallel_workers: int,
+    resume: bool,
+    dry_run: bool,
+    partition: Optional[str] = None
+) -> str:
+    """Start a migration in a background thread.
+
+    Args:
+        table: Table name (RAWDATACOR, ELABDATADISP, or all)
+        parallel_workers: Number of parallel workers (0 = sequential)
+        resume: Whether to resume from last checkpoint
+        dry_run: Whether to run in dry-run mode
+        partition: Optional partition name for single partition migration
+
+    Returns:
+        Status message
+    """
+    task_id = f"{table}_{datetime.now().strftime('%Y%m%d_%H%M%S')}"
+
+    if task_id in running_migrations:
+        return f"❌ Migration already running for {table}"
+
+    # Validate options
+    if parallel_workers > 0 and partition:
+        return "❌ Cannot use parallel workers with single partition mode"
+
+    # Initialize log buffer for this task
+    migration_logs[task_id] = []
+
+    def run_migration():
+        """Worker function to run migration."""
+        try:
+            migration_logs[task_id].append(f"Starting migration for {table}...")
+            migration_logs[task_id].append(f"Mode: {'Parallel' if parallel_workers > 0 else 'Sequential'}")
+            migration_logs[task_id].append(f"Resume: {resume}, Dry-run: {dry_run}")
+
+            if parallel_workers > 0:
+                # Parallel migration
+                rows = run_parallel_migration(
+                    table,
+                    num_workers=parallel_workers,
+                    dry_run=dry_run,
+                    resume=resume
+                )
+            else:
+                # Sequential migration
+                rows = run_full_migration(
+                    table,
+                    dry_run=dry_run,
+                    resume=resume,
+                    partition=partition
+                )
+
+            migration_logs[task_id].append(f"✓ Migration complete: {rows:,} rows migrated")
+            running_migrations[task_id] = 'completed'
+
+        except Exception as e:
+            migration_logs[task_id].append(f"❌ Migration failed: {e}")
+            running_migrations[task_id] = 'failed'
+            logger.error(f"Migration failed: {e}")
+
+    # Start migration in background thread
+    thread = threading.Thread(target=run_migration, daemon=True)
+    thread.start()
+
+    running_migrations[task_id] = 'running'
+
+    return f"✓ Migration started: {task_id}\nCheck the 'Logs' tab for progress"
+
+
+def start_incremental_migration_task(
+    table: str,
+    dry_run: bool
+) -> str:
+    """Start an incremental migration in a background thread.
+
+    Args:
+        table: Table name (RAWDATACOR, ELABDATADISP, or all)
+        dry_run: Whether to run in dry-run mode
+
+    Returns:
+        Status message
+    """
+    task_id = f"incremental_{table}_{datetime.now().strftime('%Y%m%d_%H%M%S')}"
+
+    if task_id in running_migrations:
+        return f"❌ Incremental migration already running for {table}"
+
+    # Initialize log buffer for this task
+    migration_logs[task_id] = []
+
+    def run_migration():
+        """Worker function to run incremental migration."""
+        try:
+            migration_logs[task_id].append(f"Starting INCREMENTAL migration for {table}...")
+            migration_logs[task_id].append(f"Dry-run: {dry_run}")
+            migration_logs[task_id].append("This will sync only NEW data added since last migration")
+
+            from src.migrator.incremental_migrator import run_incremental_migration
+
+            rows = run_incremental_migration(
+                table,
+                dry_run=dry_run
+            )
+
+            migration_logs[task_id].append(f"✓ Incremental migration complete: {rows:,} rows migrated")
+            running_migrations[task_id] = 'completed'
+
+        except Exception as e:
+            migration_logs[task_id].append(f"❌ Incremental migration failed: {e}")
+            running_migrations[task_id] = 'failed'
+            logger.error(f"Incremental migration failed: {e}")
+
+    # Start migration in background thread
+    thread = threading.Thread(target=run_migration, daemon=True)
+    thread.start()
+
+    running_migrations[task_id] = 'running'
+
+    return f"✓ Incremental migration started: {task_id}\nCheck the 'Logs' tab for progress"
+
+
+def get_migration_logs(task_id: str) -> str:
+    """Get logs for a specific migration task.
+
+    Args:
+        task_id: Task identifier
+
+    Returns:
+        Log contents
+    """
+    if not task_id or task_id == "No migrations running":
+        return "Select a migration to view logs"
+
+    logs = migration_logs.get(task_id, [])
+    if not logs:
+        return f"No logs available for {task_id}"
+
+    return "\n".join(logs)
+
+
+def refresh_migration_state():
+    """Refresh migration state display."""
+    return get_migration_state_df()
+
+
+def refresh_status_chart():
+    """Refresh status chart."""
+    return create_status_chart()
+
+
+def refresh_running_migrations() -> gr.Dropdown:
+    """Refresh list of running migrations."""
+    if not running_migrations:
+        return gr.Dropdown(choices=["No migrations running"], value="No migrations running")
+
+    choices = list(running_migrations.keys())
+    return gr.Dropdown(choices=choices, value=choices[0] if choices else None)
+
+
+# Build Gradio interface
+with gr.Blocks(title="MySQL to PostgreSQL Migration", theme=gr.themes.Soft()) as demo:
+    gr.Markdown("# 🔄 MySQL to PostgreSQL Migration Dashboard")
+    gr.Markdown("Monitor and control data migration from MySQL to PostgreSQL")
+
+    with gr.Tabs():
+        # Tab 1: Overview
+        with gr.Tab("📊 Overview"):
+            gr.Markdown("## Migration Status Overview")
+
+            with gr.Row():
+                refresh_btn = gr.Button("🔄 Refresh", scale=1)
+
+            with gr.Row():
+                status_chart = gr.Plot(label="Partition Status by Table")
+
+            with gr.Row():
+                state_table = gr.Dataframe(
+                    value=get_migration_state_df(),
+                    label="Migration State (All Partitions)",
+                    interactive=False,
+                    wrap=True
+                )
+
+            # Auto-refresh every 10 seconds
+            refresh_btn.click(
+                fn=lambda: (refresh_migration_state(), refresh_status_chart()),
+                outputs=[state_table, status_chart]
+            )
+
+            # Initial load
+            demo.load(
+                fn=lambda: (refresh_migration_state(), refresh_status_chart()),
+                outputs=[state_table, status_chart]
+            )
+
+        # Tab 2: Start Migration
+        with gr.Tab("▶️ Start Migration"):
+            gr.Markdown("## Start New Migration")
+
+            with gr.Row():
+                table_select = gr.Dropdown(
+                    choices=["RAWDATACOR", "ELABDATADISP", "all"],
+                    value="RAWDATACOR",
+                    label="Table to Migrate"
+                )
+
+                partition_input = gr.Textbox(
+                    label="Partition (optional)",
+                    placeholder="Leave empty for all partitions, or enter e.g. 'part7', 'd9'",
+                    value=""
+                )
+
+            with gr.Row():
+                parallel_slider = gr.Slider(
+                    minimum=0,
+                    maximum=10,
+                    value=0,
+                    step=1,
+                    label="Parallel Workers (0 = sequential)"
+                )
+
+            with gr.Row():
+                resume_check = gr.Checkbox(label="Resume from last checkpoint", value=True)
+                dry_run_check = gr.Checkbox(label="Dry run (simulate without writing)", value=False)
+
+            start_btn = gr.Button("▶️ Start Migration", variant="primary", size="lg")
+
+            migration_output = gr.Textbox(
+                label="Migration Status",
+                lines=3,
+                interactive=False
+            )
+
+            start_btn.click(
+                fn=start_migration_task,
+                inputs=[table_select, parallel_slider, resume_check, dry_run_check, partition_input],
+                outputs=migration_output
+            )
+
+        # Tab 3: Incremental Migration
+        with gr.Tab("🔄 Incremental Sync"):
+            gr.Markdown("## Incremental Migration (Sync New Data)")
+            gr.Markdown("""
+            Questa modalità sincronizza **solo i dati nuovi** aggiunti dopo l'ultima migrazione full.
+
+            **Come funziona:**
+            - Trova le nuove consolidation keys in MySQL che non esistono ancora in PostgreSQL
+            - Migra solo quelle chiavi (non riprocessa dati già migrati)
+            - Usa `migration_state` per tracciare l'ultima chiave processata
+
+            **Quando usare:**
+            - Dopo aver completato una migrazione full di tutte le partizioni
+            - Per sincronizzare periodicamente nuovi dati senza rifare tutto
+            - Per aggiornamenti quotidiani/settimanali
+            """)
+
+            with gr.Row():
+                inc_table_select = gr.Dropdown(
+                    choices=["RAWDATACOR", "ELABDATADISP", "all"],
+                    value="RAWDATACOR",
+                    label="Table to Sync"
+                )
+
+            with gr.Row():
+                inc_dry_run_check = gr.Checkbox(
+                    label="Dry run (simulate without writing)",
+                    value=False
+                )
+
+            inc_start_btn = gr.Button("🔄 Start Incremental Sync", variant="primary", size="lg")
+
+            inc_migration_output = gr.Textbox(
+                label="Sync Status",
+                lines=3,
+                interactive=False
+            )
+
+            inc_start_btn.click(
+                fn=start_incremental_migration_task,
+                inputs=[inc_table_select, inc_dry_run_check],
+                outputs=inc_migration_output
+            )
+
+        # Tab 4: Logs
+        with gr.Tab("📝 Logs"):
+            gr.Markdown("## Migration Logs")
+
+            with gr.Row():
+                running_migrations_dropdown = gr.Dropdown(
+                    choices=["No migrations running"],
+                    value="No migrations running",
+                    label="Select Migration",
+                    interactive=True
+                )
+                refresh_migrations_btn = gr.Button("🔄 Refresh List")
+
+            log_output = gr.Textbox(
+                label="Log Output",
+                lines=20,
+                interactive=False,
+                max_lines=50
+            )
+
+            refresh_migrations_btn.click(
+                fn=refresh_running_migrations,
+                outputs=running_migrations_dropdown
+            )
+
+            running_migrations_dropdown.change(
+                fn=get_migration_logs,
+                inputs=running_migrations_dropdown,
+                outputs=log_output
+            )
+
+        # Tab 5: Error Logs
+        with gr.Tab("⚠️ Error Logs"):
+            gr.Markdown("## Error Log Viewer")
+            gr.Markdown("View logged validation errors (invalid consolidation keys)")
+
+            with gr.Row():
+                error_log_dropdown = gr.Dropdown(
+                    choices=[x[0] for x in get_error_logs()] if get_error_logs() else ["No error logs found"],
+                    label="Select Error Log File",
+                    value="Select a log file..."
+                )
+                refresh_error_logs_btn = gr.Button("🔄 Refresh")
+
+            error_log_content = gr.Textbox(
+                label="Error Log Contents",
+                lines=20,
+                interactive=False,
+                max_lines=50
+            )
+
+            def refresh_error_log_list():
+                logs = get_error_logs()
+                choices = [x[0] for x in logs] if logs else ["No error logs found"]
+                return gr.Dropdown(choices=choices)
+
+            def show_error_log(filename):
+                if not filename or filename == "Select a log file..." or filename == "No error logs found":
+                    return "Select a log file to view its contents"
+                return read_error_log(filename)
+
+            refresh_error_logs_btn.click(
+                fn=refresh_error_log_list,
+                outputs=error_log_dropdown
+            )
+
+            error_log_dropdown.change(
+                fn=show_error_log,
+                inputs=error_log_dropdown,
+                outputs=error_log_content
+            )
+
+
+def launch_ui(share=False, server_port=7860):
+    """Launch the Gradio UI.
+
+    Args:
+        share: Whether to create a public share link
+        server_port: Port to run the server on
+    """
+    demo.launch(
+        share=share,
+        server_port=server_port,
+        server_name="0.0.0.0"  # Listen on all interfaces
+    )
+
+
+if __name__ == "__main__":
+    import sys
+
+    # Parse command line args
+    share = "--share" in sys.argv
+    port = 7860
+
+    for arg in sys.argv:
+        if arg.startswith("--port="):
+            port = int(arg.split("=")[1])
+
+    print(f"Starting Migration Dashboard on port {port}...")
+    print(f"Share mode: {share}")
+
+    launch_ui(share=share, server_port=port)