# Ledger Backfill Strategy

**Version**: 1.0.0  
**Last Updated**: 2025-01-20  
**Status**: Active Documentation

## Overview

This document outlines the strategy for backfilling historical ledger data into the DBIS Core Banking System ledger. The backfill process ensures data integrity, maintains idempotency, and supports resumable operations.

---

## Backfill Scenarios

### Scenario 1: Initial System Setup (Empty Ledger)

**Use Case**: Setting up a new DBIS instance with empty ledger, populating from external source (e.g., legacy system, CSV export, external API).

**Approach**:
1. Validate source data integrity
2. Transform source data to DBIS ledger format
3. Batch insert with idempotency checks
4. Verify balance consistency
5. Apply constraints after backfill

### Scenario 2: Schema Migration (Existing Ledger Data)

**Use Case**: Migrating existing ledger data to new schema (e.g., adding new fields, restructuring).

**Approach**:
1. Audit existing data
2. Transform to new schema format
3. Migrate in batches
4. Verify data integrity
5. Update schema constraints

### Scenario 3: Data Reconciliation (Fix Inconsistencies)

**Use Case**: Fixing inconsistent balances or missing entries discovered during audit.

**Approach**:
1. Identify inconsistencies
2. Generate correction entries
3. Apply corrections via normal posting function
4. Verify balance consistency
5. Document corrections in audit log

### Scenario 4: Dual-Ledger Sync (SCB Ledger Backfill)

**Use Case**: Backfilling historical entries from SCB (Sovereign Central Bank) ledger to DBIS.

**Approach**:
1. Extract entries from SCB ledger
2. Transform to DBIS format
3. Post to DBIS via outbox pattern
4. Track sync status
5. Verify dual-ledger consistency

---

## Backfill Architecture

### Component Overview

```
┌─────────────────────────────────────────────────────────────┐
│                    Backfill Architecture                     │
├─────────────────────────────────────────────────────────────┤
│                                                             │
│  ┌─────────────┐     ┌──────────────┐     ┌─────────────┐ │
│  │   Source    │────▶│   Transform  │────▶│   Validate  │ │
│  │   Reader    │     │    Service   │     │   Service   │ │
│  └─────────────┘     └──────────────┘     └─────────────┘ │
│                                                             │
│                            │                                │
│                            ▼                                │
│                    ┌──────────────┐                         │
│                    │    Batch     │                         │
│                    │   Processor  │                         │
│                    └──────────────┘                         │
│                            │                                │
│                ┌───────────┴───────────┐                    │
│                ▼                       ▼                    │
│        ┌──────────────┐       ┌──────────────┐             │
│        │   Ledger     │       │   Checkpoint │             │
│        │   Posting    │       │    Service   │             │
│        │   Module     │       └──────────────┘             │
│        └──────────────┘                                    │
│                │                                            │
│                ▼                                            │
│        ┌──────────────┐                                    │
│        │   Audit &    │                                    │
│        │ Verification │                                    │
│        └──────────────┘                                    │
│                                                             │
└─────────────────────────────────────────────────────────────┘
```

### Key Components

1. **Source Reader**: Reads data from source (CSV, API, database, etc.)
2. **Transform Service**: Transforms source data to DBIS ledger format
3. **Validate Service**: Validates entries before posting
4. **Batch Processor**: Processes entries in batches with checkpointing
5. **Ledger Posting Module**: Uses atomic posting function for entries
6. **Checkpoint Service**: Tracks progress for resumable backfill
7. **Audit & Verification**: Validates backfill results

---

## Backfill Process

### Step 1: Pre-Backfill Preparation

#### 1.1 Audit Existing Data

Before starting backfill, audit existing data:

```sql
-- Check for existing ledger entries
SELECT COUNT(*), MIN(timestamp_utc), MAX(timestamp_utc)
FROM ledger_entries;

-- Check for inconsistent balances
SELECT id, balance, available_balance, reserved_balance
FROM bank_accounts
WHERE available_balance < 0
   OR reserved_balance < 0
   OR available_balance > balance
   OR (available_balance + reserved_balance) > balance;

-- Check for duplicate reference IDs
SELECT ledger_id, reference_id, COUNT(*)
FROM ledger_entries
GROUP BY ledger_id, reference_id
HAVING COUNT(*) > 1;
```

#### 1.2 Verify Schema

Ensure all required migrations are applied:

```sql
-- Verify idempotency constraint exists
SELECT constraint_name 
FROM information_schema.table_constraints 
WHERE table_name = 'ledger_entries' 
  AND constraint_name LIKE '%reference%';

-- Verify outbox table exists
SELECT COUNT(*) FROM dual_ledger_outbox;

-- Verify posting function exists
SELECT routine_name 
FROM information_schema.routines 
WHERE routine_name = 'post_ledger_entry';
```

#### 1.3 Prepare Source Data

- **CSV Export**: Ensure format matches expected schema
- **API Extraction**: Configure API endpoints and authentication
- **Database Extraction**: Set up connection and query filters
- **Legacy System**: Configure export format and mapping

---

### Step 2: Data Transformation

#### 2.1 Source Data Format

Source data should be transformed to this format:

```typescript
interface LedgerEntrySource {
  ledgerId: string;           // e.g., "MASTER", "SOVEREIGN"
  debitAccountId: string;     // Account ID
  creditAccountId: string;    // Account ID
  amount: string;             // Decimal as string (e.g., "1000.00")
  currencyCode: string;       // ISO 4217 (e.g., "USD")
  assetType: string;          // "fiat", "cbdc", "commodity", "security"
  transactionType: string;    // Transaction type classification
  referenceId: string;        // Unique reference ID (required for idempotency)
  timestampUtc?: string;      // ISO 8601 timestamp
  fxRate?: string;            // FX rate if applicable
  metadata?: Record<string, unknown>; // Additional metadata
}
```

#### 2.2 Transformation Rules

1. **Account ID Mapping**: Map source account identifiers to DBIS account IDs
2. **Amount Normalization**: Convert amounts to standard format (decimal string)
3. **Currency Validation**: Validate currency codes against ISO 4217
4. **Timestamp Normalization**: Convert timestamps to UTC ISO 8601 format
5. **Reference ID Generation**: Generate unique reference IDs if not present
6. **Metadata Extraction**: Extract relevant metadata from source

#### 2.3 Example Transformation Script

```typescript
// Example: Transform CSV data
function transformCSVToLedgerEntry(csvRow: CSVRow): LedgerEntrySource {
  return {
    ledgerId: csvRow.ledger || 'MASTER',
    debitAccountId: mapAccountId(csvRow.debit_account),
    creditAccountId: mapAccountId(csvRow.credit_account),
    amount: normalizeAmount(csvRow.amount),
    currencyCode: csvRow.currency || 'USD',
    assetType: csvRow.asset_type || 'fiat',
    transactionType: mapTransactionType(csvRow.txn_type),
    referenceId: csvRow.reference_id || generateReferenceId(csvRow),
    timestampUtc: csvRow.timestamp || new Date().toISOString(),
    fxRate: csvRow.fx_rate || undefined,
    metadata: extractMetadata(csvRow),
  };
}
```

---

### Step 3: Batch Processing

#### 3.1 Batch Configuration

Configure batch processing parameters:

```typescript
interface BackfillConfig {
  batchSize: number;           // Entries per batch (default: 1000)
  checkpointInterval: number;  // Checkpoint every N batches (default: 10)
  maxRetries: number;          // Max retries per batch (default: 3)
  retryDelay: number;          // Initial retry delay in ms (default: 1000)
  parallelWorkers: number;     // Number of parallel workers (default: 1)
  skipDuplicates: boolean;     // Skip entries with duplicate reference IDs (default: true)
  validateBalances: boolean;   // Validate balances after each batch (default: true)
}
```

#### 3.2 Checkpointing Strategy

Use checkpointing to enable resumable backfill:

```sql
-- Create checkpoint table for ledger backfill
CREATE TABLE IF NOT EXISTS ledger_backfill_checkpoints (
  id SERIAL PRIMARY KEY,
  source_id VARCHAR(255) NOT NULL,
  source_type VARCHAR(50) NOT NULL, -- 'CSV', 'API', 'DATABASE', 'SCB'
  last_processed_id VARCHAR(255),
  last_processed_timestamp TIMESTAMP,
  total_processed BIGINT DEFAULT 0,
  total_successful BIGINT DEFAULT 0,
  total_failed BIGINT DEFAULT 0,
  status VARCHAR(50) DEFAULT 'IN_PROGRESS', -- 'IN_PROGRESS', 'COMPLETED', 'FAILED', 'PAUSED'
  started_at TIMESTAMP DEFAULT NOW(),
  last_checkpoint_at TIMESTAMP DEFAULT NOW(),
  completed_at TIMESTAMP,
  error_message TEXT,
  metadata JSONB,
  UNIQUE(source_id, source_type)
);

CREATE INDEX idx_backfill_checkpoints_status 
ON ledger_backfill_checkpoints(status);

CREATE INDEX idx_backfill_checkpoints_source 
ON ledger_backfill_checkpoints(source_id, source_type);
```

#### 3.3 Batch Processing Loop

```typescript
async function processBackfill(
  source: DataSource,
  config: BackfillConfig
): Promise<BackfillResult> {
  const checkpoint = await loadCheckpoint(source.id, source.type);
  let processed = 0;
  let successful = 0;
  let failed = 0;
  let lastProcessedId: string | null = null;
  let lastProcessedTimestamp: Date | null = null;

  while (true) {
    // Load batch from source (starting from checkpoint)
    const batch = await source.readBatch({
      startId: checkpoint?.lastProcessedId,
      startTimestamp: checkpoint?.lastProcessedTimestamp,
      limit: config.batchSize,
    });

    if (batch.length === 0) {
      break; // No more data
    }

    // Process batch
    const results = await processBatch(batch, config);

    // Update counters
    processed += batch.length;
    successful += results.successful;
    failed += results.failed;

    // Update checkpoint
    lastProcessedId = batch[batch.length - 1].id;
    lastProcessedTimestamp = batch[batch.length - 1].timestamp;

    await saveCheckpoint({
      sourceId: source.id,
      sourceType: source.type,
      lastProcessedId,
      lastProcessedTimestamp,
      totalProcessed: processed,
      totalSuccessful: successful,
      totalFailed: failed,
      status: 'IN_PROGRESS',
    });

    // Validate balances if configured
    if (config.validateBalances && processed % (config.checkpointInterval * config.batchSize) === 0) {
      await validateBalances();
    }
  }

  // Mark as completed
  await saveCheckpoint({
    sourceId: source.id,
    sourceType: source.type,
    status: 'COMPLETED',
    completedAt: new Date(),
  });

  return {
    totalProcessed: processed,
    totalSuccessful: successful,
    totalFailed: failed,
  };
}
```

---

### Step 4: Entry Posting

#### 4.1 Use Atomic Posting Function

Always use the atomic posting function for backfill entries:

```typescript
async function postBackfillEntry(entry: LedgerEntrySource): Promise<void> {
  try {
    // Use atomic posting function via SQL
    const result = await prisma.$executeRaw`
      SELECT post_ledger_entry(
        ${entry.ledgerId}::TEXT,
        ${entry.debitAccountId}::TEXT,
        ${entry.creditAccountId}::TEXT,
        ${entry.amount}::NUMERIC,
        ${entry.currencyCode}::TEXT,
        ${entry.assetType}::TEXT,
        ${entry.transactionType}::TEXT,
        ${entry.referenceId}::TEXT,
        ${entry.fxRate || null}::NUMERIC,
        ${entry.metadata ? JSON.stringify(entry.metadata) : null}::JSONB
      )
    `;

    // Verify result
    if (!result) {
      throw new Error('Failed to post ledger entry');
    }
  } catch (error) {
    // Handle idempotency violation (duplicate reference ID)
    if (error.code === '23505' || error.message?.includes('duplicate')) {
      // Skip duplicate entries if configured
      if (config.skipDuplicates) {
        return; // Entry already exists, skip
      }
      throw new Error(`Duplicate reference ID: ${entry.referenceId}`);
    }

    throw error;
  }
}
```

#### 4.2 Batch Posting

Post entries in batches for efficiency:

```typescript
async function processBatch(
  entries: LedgerEntrySource[],
  config: BackfillConfig
): Promise<{ successful: number; failed: number }> {
  let successful = 0;
  let failed = 0;

  // Process in parallel if configured
  if (config.parallelWorkers > 1) {
    const chunks = chunkArray(entries, config.parallelWorkers);
    const results = await Promise.allSettled(
      chunks.map((chunk) => processChunk(chunk, config))
    );

    for (const result of results) {
      if (result.status === 'fulfilled') {
        successful += result.value.successful;
        failed += result.value.failed;
      } else {
        failed += entries.length;
      }
    }
  } else {
    // Sequential processing
    for (const entry of entries) {
      try {
        await postBackfillEntry(entry);
        successful++;
      } catch (error) {
        failed++;
        logError(entry, error);
      }
    }
  }

  return { successful, failed };
}
```

---

### Step 5: Balance Constraints Application

#### 5.1 Pre-Constraint Validation

Before applying balance constraints, validate all balances:

```sql
-- Validate all balances are consistent
DO $$
DECLARE
  inconsistent_count INTEGER;
BEGIN
  SELECT COUNT(*) INTO inconsistent_count
  FROM bank_accounts
  WHERE available_balance < 0
     OR reserved_balance < 0
     OR available_balance > balance
     OR (available_balance + reserved_balance) > balance;

  IF inconsistent_count > 0 THEN
    RAISE EXCEPTION 'Found % inconsistent balances. Fix before applying constraints.', inconsistent_count;
  END IF;
END $$;
```

#### 5.2 Apply Constraints

After backfill completes and balances are validated, apply constraints:

```bash
# Apply balance constraints migration
psql $DATABASE_URL -f db/migrations/004_balance_constraints.sql
```

#### 5.3 Post-Constraint Verification

Verify constraints are applied correctly:

```sql
-- Check constraint exists
SELECT constraint_name, constraint_type
FROM information_schema.table_constraints
WHERE table_name = 'bank_accounts'
  AND constraint_name LIKE '%balance%';

-- Verify constraint is enforced
-- This should fail if constraints are working
UPDATE bank_accounts
SET available_balance = -1
WHERE id = (SELECT id FROM bank_accounts LIMIT 1);
```

---

### Step 6: Verification and Reconciliation

#### 6.1 Entry Verification

Verify all entries were posted correctly:

```sql
-- Compare source count vs. posted count
SELECT 
  COUNT(*) as total_entries,
  COUNT(DISTINCT reference_id) as unique_references,
  COUNT(DISTINCT ledger_id) as unique_ledgers,
  MIN(timestamp_utc) as earliest_entry,
  MAX(timestamp_utc) as latest_entry
FROM ledger_entries
WHERE reference_id LIKE 'BACKFILL-%'; -- If using prefix for backfill entries

-- Check for missing entries
SELECT source_id, reference_id
FROM backfill_source_data
WHERE NOT EXISTS (
  SELECT 1 FROM ledger_entries
  WHERE reference_id = backfill_source_data.reference_id
);
```

#### 6.2 Balance Reconciliation

Reconcile balances after backfill:

```sql
-- Compare expected vs. actual balances
SELECT 
  account_id,
  expected_balance,
  actual_balance,
  (expected_balance - actual_balance) as difference
FROM (
  SELECT 
    account_id,
    SUM(CASE WHEN side = 'debit' THEN amount ELSE -amount END) as expected_balance,
    (SELECT balance FROM bank_accounts WHERE id = account_id) as actual_balance
  FROM ledger_entries
  WHERE account_id IN (SELECT id FROM bank_accounts)
  GROUP BY account_id
) reconciliation
WHERE ABS(expected_balance - actual_balance) > 0.01; -- Allow small rounding differences
```

#### 6.3 Dual-Ledger Reconciliation

If backfilling from SCB ledger, reconcile dual-ledger consistency:

```sql
-- Check outbox sync status
SELECT 
  status,
  COUNT(*) as count,
  MIN(created_at) as oldest,
  MAX(created_at) as newest
FROM dual_ledger_outbox
WHERE created_at >= (SELECT MIN(timestamp_utc) FROM ledger_entries WHERE reference_id LIKE 'BACKFILL-%')
GROUP BY status;

-- Verify all entries have corresponding outbox records (for SCB sync)
SELECT le.id, le.reference_id
FROM ledger_entries le
WHERE le.reference_id LIKE 'BACKFILL-%'
  AND NOT EXISTS (
    SELECT 1 FROM dual_ledger_outbox dlo
    WHERE dlo.reference_id = le.reference_id
  );
```

---

## Implementation Scripts

### TypeScript Backfill Script

**File**: `dbis_core/scripts/backfill-ledger.ts`

```typescript
#!/usr/bin/env ts-node
import { PrismaClient } from '@prisma/client';
import { readFileSync } from 'fs';
import { parse } from 'csv-parse/sync';

const prisma = new PrismaClient();

interface BackfillConfig {
  sourceFile: string;
  ledgerId: string;
  batchSize: number;
  skipDuplicates: boolean;
}

async function backfillFromCSV(config: BackfillConfig) {
  // Read and parse CSV
  const csvData = readFileSync(config.sourceFile, 'utf-8');
  const records = parse(csvData, {
    columns: true,
    skip_empty_lines: true,
  });

  let processed = 0;
  let successful = 0;
  let failed = 0;

  // Process in batches
  for (let i = 0; i < records.length; i += config.batchSize) {
    const batch = records.slice(i, i + config.batchSize);
    
    for (const record of batch) {
      try {
        // Transform and post entry
        await prisma.$executeRaw`
          SELECT post_ledger_entry(
            ${config.ledgerId}::TEXT,
            ${record.debitAccountId}::TEXT,
            ${record.creditAccountId}::TEXT,
            ${record.amount}::NUMERIC,
            ${record.currencyCode}::TEXT,
            ${record.assetType || 'fiat'}::TEXT,
            ${record.transactionType}::TEXT,
            ${record.referenceId}::TEXT,
            ${record.fxRate || null}::NUMERIC,
            ${record.metadata ? JSON.stringify(JSON.parse(record.metadata)) : null}::JSONB
          )
        `;
        successful++;
      } catch (error) {
        if (config.skipDuplicates && error.code === '23505') {
          // Skip duplicates
          continue;
        }
        failed++;
        console.error(`Failed to post entry ${record.referenceId}:`, error.message);
      }
      processed++;
    }

    console.log(`Processed ${processed}/${records.length} entries (${successful} successful, ${failed} failed)`);
  }

  return { processed, successful, failed };
}

// CLI entry point
const config: BackfillConfig = {
  sourceFile: process.env.BACKFILL_SOURCE_FILE || 'backfill.csv',
  ledgerId: process.env.LEDGER_ID || 'MASTER',
  batchSize: parseInt(process.env.BATCH_SIZE || '1000', 10),
  skipDuplicates: process.env.SKIP_DUPLICATES === 'true',
};

backfillFromCSV(config)
  .then((result) => {
    console.log('Backfill completed:', result);
    process.exit(0);
  })
  .catch((error) => {
    console.error('Backfill failed:', error);
    process.exit(1);
  })
  .finally(() => {
    prisma.$disconnect();
  });
```

### SQL Backfill Script

**File**: `dbis_core/scripts/backfill-ledger.sql`

```sql
-- Ledger Backfill Script
-- Use this for direct SQL-based backfill from another database

-- Example: Backfill from external ledger_entries_legacy table
DO $$
DECLARE
  batch_size INTEGER := 1000;
  processed INTEGER := 0;
  successful INTEGER := 0;
  failed INTEGER := 0;
  entry RECORD;
BEGIN
  -- Create temporary table for batch processing
  CREATE TEMP TABLE backfill_batch AS
  SELECT * FROM ledger_entries_legacy
  ORDER BY id
  LIMIT 0;

  -- Process in batches
  FOR entry IN 
    SELECT * FROM ledger_entries_legacy
    ORDER BY id
  LOOP
    BEGIN
      -- Post entry using atomic function
      PERFORM post_ledger_entry(
        entry.ledger_id::TEXT,
        entry.debit_account_id::TEXT,
        entry.credit_account_id::TEXT,
        entry.amount::NUMERIC,
        entry.currency_code::TEXT,
        entry.asset_type::TEXT,
        entry.transaction_type::TEXT,
        entry.reference_id::TEXT,
        entry.fx_rate::NUMERIC,
        entry.metadata::JSONB
      );
      
      successful := successful + 1;
    EXCEPTION
      WHEN unique_violation THEN
        -- Duplicate reference ID, skip if configured
        failed := failed + 1;
        RAISE NOTICE 'Skipping duplicate reference ID: %', entry.reference_id;
      WHEN OTHERS THEN
        failed := failed + 1;
        RAISE NOTICE 'Error processing entry %: %', entry.reference_id, SQLERRM;
    END;

    processed := processed + 1;

    -- Checkpoint every batch_size entries
    IF processed % batch_size = 0 THEN
      RAISE NOTICE 'Processed % entries (% successful, % failed)', processed, successful, failed;
    END IF;
  END LOOP;

  RAISE NOTICE 'Backfill completed: % total, % successful, % failed', processed, successful, failed;
END $$;
```

---

## Best Practices

### 1. Idempotency

- Always use unique `reference_id` for each entry
- Use atomic posting function that enforces idempotency
- Skip duplicates during backfill if they already exist

### 2. Checkpointing

- Save checkpoint after each batch
- Enable resumable backfill from last checkpoint
- Track progress with metrics (processed, successful, failed)

### 3. Validation

- Validate source data before transformation
- Validate transformed entries before posting
- Verify balances after backfill completion
- Reconcile with source system if possible

### 4. Error Handling

- Log all errors with full context
- Retry transient errors with exponential backoff
- Skip permanent errors (e.g., duplicate reference IDs)
- Generate error report after completion

### 5. Performance

- Process in batches (1000-10000 entries per batch)
- Use parallel processing for large backfills
- Monitor database performance during backfill
- Schedule during low-traffic periods

### 6. Testing

- Test backfill process on staging environment first
- Use small test dataset to verify transformation
- Verify balances match expected values
- Test rollback procedures if needed

---

## Monitoring and Metrics

### Key Metrics to Track

1. **Progress Metrics**:
   - Total entries to process
   - Entries processed
   - Entries successful
   - Entries failed
   - Processing rate (entries/second)

2. **Performance Metrics**:
   - Batch processing time
   - Database query time
   - Checkpoint save time
   - Total elapsed time

3. **Quality Metrics**:
   - Duplicate entries skipped
   - Validation errors
   - Balance inconsistencies
   - Reconciliation mismatches

### Monitoring Queries

```sql
-- Check backfill progress
SELECT 
  source_id,
  source_type,
  status,
  total_processed,
  total_successful,
  total_failed,
  last_checkpoint_at,
  NOW() - last_checkpoint_at as time_since_last_checkpoint
FROM ledger_backfill_checkpoints
WHERE status = 'IN_PROGRESS';

-- Check for stalled backfills
SELECT *
FROM ledger_backfill_checkpoints
WHERE status = 'IN_PROGRESS'
  AND last_checkpoint_at < NOW() - INTERVAL '1 hour';

-- Verify backfill completion
SELECT 
  COUNT(*) as total_entries,
  MIN(timestamp_utc) as earliest,
  MAX(timestamp_utc) as latest
FROM ledger_entries
WHERE reference_id LIKE 'BACKFILL-%';
```

---

## Rollback Procedures

### Scenario 1: Rollback Before Constraints Applied

If constraints have not been applied, rollback is straightforward:

```sql
-- Remove backfilled entries
DELETE FROM ledger_entries
WHERE reference_id LIKE 'BACKFILL-%';

-- Remove outbox records
DELETE FROM dual_ledger_outbox
WHERE reference_id LIKE 'BACKFILL-%';

-- Reset balances (if needed)
UPDATE bank_accounts
SET balance = balance - (
  SELECT COALESCE(SUM(amount), 0)
  FROM ledger_entries
  WHERE debit_account_id = bank_accounts.id
    AND reference_id LIKE 'BACKFILL-%'
) + (
  SELECT COALESCE(SUM(amount), 0)
  FROM ledger_entries
  WHERE credit_account_id = bank_accounts.id
    AND reference_id LIKE 'BACKFILL-%'
);
```

### Scenario 2: Rollback After Constraints Applied

If constraints have been applied, rollback is more complex:

1. Temporarily disable constraints
2. Remove backfilled entries
3. Recalculate balances
4. Re-enable constraints
5. Verify balance consistency

**Note**: This should only be done during maintenance window.

---

## Troubleshooting

### Common Issues

#### 1. Duplicate Reference ID Errors

**Symptom**: `unique_violation` error on `reference_id`

**Solution**:
- Check if entries were already backfilled
- Use `skipDuplicates: true` to skip existing entries
- Or regenerate reference IDs for duplicates

#### 2. Balance Inconsistencies

**Symptom**: Balance validation fails

**Solution**:
- Identify accounts with inconsistent balances
- Generate correction entries
- Post corrections before applying constraints

#### 3. Slow Performance

**Symptom**: Backfill processing is slow

**Solution**:
- Increase batch size (if memory allows)
- Use parallel processing
- Optimize database indexes
- Run during off-peak hours

#### 4. Out of Memory

**Symptom**: Process runs out of memory

**Solution**:
- Reduce batch size
- Process sequentially instead of parallel
- Use streaming instead of loading all data

---

## Examples

### Example 1: CSV Backfill

```bash
# Configure environment
export DATABASE_URL="postgresql://user:password@host:port/database"
export BACKFILL_SOURCE_FILE="ledger_export.csv"
export LEDGER_ID="MASTER"
export BATCH_SIZE="1000"
export SKIP_DUPLICATES="true"

# Run backfill script
cd dbis_core
ts-node scripts/backfill-ledger.ts
```

### Example 2: SCB Ledger Sync

```typescript
// Backfill from SCB ledger via API
async function backfillFromSCB(sovereignBankId: string, startDate: Date, endDate: Date) {
  const scbApi = new SCBLedgerAPI(sovereignBankId);
  const entries = await scbApi.getLedgerEntries(startDate, endDate);

  for (const entry of entries) {
    // Transform SCB entry to DBIS format
    const dbisEntry = transformSCBEntry(entry);

    // Post to DBIS (will create outbox record for dual-ledger sync)
    await ledgerPostingModule.postEntry(dbisEntry);
  }
}
```

---

## References

- Migration Files: `dbis_core/db/migrations/`
- Ledger Posting Module: `dbis_core/src/core/ledger/ledger-posting.module.ts`
- Atomic Posting Function: `dbis_core/db/migrations/005_post_ledger_entry.sql`
- Block Indexer Backfill: `explorer-monorepo/backend/indexer/backfill/backfill.go` (reference pattern)