stella-ops.org/git.stella-ops.org
1
0
Fork 0
Code Issues Pull Requests Actions 2 Releases Wiki Activity

fix(jobengine): make all orchestrator migration SQL idempotent and PostgreSQL-compatible

Browse Source 
Fix 4 classes of issues that prevented JobEngine from auto-migrating:
1. Non-idempotent DDL: add IF NOT EXISTS to CREATE TABLE, wrap CREATE
   TYPE in DO blocks with EXCEPTION WHEN duplicate_object, wrap partition
   creation with EXCEPTION WHEN duplicate_object OR SQLSTATE '42P17'
2. Reserved keyword: quote `window` column name in 004_slo_quotas.sql
3. Invalid syntax: replace DELETE...LIMIT with ctid subquery pattern
   in 004_slo_quotas.sql and 005_audit_ledger.sql
4. Partition constraint: add tenant_id to UNIQUE(log_id) constraint
   on pack_run_logs in 006_pack_runs.sql (partitioned tables require
   partition key in all unique constraints)
5. Non-immutable index predicate: remove NOW() from partial index
   predicate in 002_backfill.sql
6. Remove BEGIN/COMMIT wrappers from all migration files (the
   StartupMigrationHost already wraps each migration in a transaction)

All 8 orchestrator migrations (001-008) now apply cleanly on fresh DB.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
master
2026-03-09 08:38:20 +02:00
parent 71db8d4386
commit 1e53976ffb
8 changed files with 266 additions and 208 deletions
Download Patch File Download Diff File Expand all files Collapse all files

67
src/JobEngine/StellaOps.JobEngine/StellaOps.JobEngine.Infrastructure/migrations/002_backfill.sql
View File

@@ -1,22 +1,23 @@
-- 002_backfill.sql
-- Backfill and watermark tables for event-time window tracking (ORCH-SVC-33-003)
-- Adds watermarks, backfill_requests, and processed_events for duplicate suppression.
BEGIN;
-- All statements are idempotent so the migration is safe on pre-existing databases.
-- Backfill request status
CREATE TYPE backfill_status AS ENUM (
'pending',
'validating',
'running',
'paused',
'completed',
'failed',
'canceled'
);
DO $$ BEGIN
CREATE TYPE backfill_status AS ENUM (
'pending',
'validating',
'running',
'paused',
'completed',
'failed',
'canceled'
);
EXCEPTION WHEN duplicate_object OR SQLSTATE '42P17' THEN NULL; END $$;
-- Watermarks: Per-source/job-type event-time cursors
CREATE TABLE watermarks (
CREATE TABLE IF NOT EXISTS watermarks (
watermark_id UUID NOT NULL,
tenant_id TEXT NOT NULL,
source_id UUID,
@@ -35,13 +36,15 @@ CREATE TABLE watermarks (
CONSTRAINT ck_watermarks_hash_hex CHECK (last_batch_hash IS NULL OR last_batch_hash ~ '^[0-9a-f]{64}$')
) PARTITION BY LIST (tenant_id);
CREATE TABLE watermarks_default PARTITION OF watermarks DEFAULT;
DO $$ BEGIN
CREATE TABLE watermarks_default PARTITION OF watermarks DEFAULT;
EXCEPTION WHEN duplicate_object OR SQLSTATE '42P17' THEN NULL; END $$;
CREATE INDEX ix_watermarks_source ON watermarks (tenant_id, source_id) WHERE source_id IS NOT NULL;
CREATE INDEX ix_watermarks_job_type ON watermarks (tenant_id, job_type) WHERE job_type IS NOT NULL;
CREATE INDEX IF NOT EXISTS ix_watermarks_source ON watermarks (tenant_id, source_id) WHERE source_id IS NOT NULL;
CREATE INDEX IF NOT EXISTS ix_watermarks_job_type ON watermarks (tenant_id, job_type) WHERE job_type IS NOT NULL;
-- Backfill Requests: Batch reprocessing operations
CREATE TABLE backfill_requests (
CREATE TABLE IF NOT EXISTS backfill_requests (
backfill_id UUID NOT NULL,
tenant_id TEXT NOT NULL,
source_id UUID,
@@ -79,14 +82,16 @@ CREATE TABLE backfill_requests (
CONSTRAINT ck_backfill_batch_size CHECK (batch_size > 0 AND batch_size <= 10000)
) PARTITION BY LIST (tenant_id);
CREATE TABLE backfill_requests_default PARTITION OF backfill_requests DEFAULT;
DO $$ BEGIN
CREATE TABLE backfill_requests_default PARTITION OF backfill_requests DEFAULT;
EXCEPTION WHEN duplicate_object OR SQLSTATE '42P17' THEN NULL; END $$;
CREATE INDEX ix_backfill_status ON backfill_requests (tenant_id, status, created_at DESC);
CREATE INDEX ix_backfill_scope ON backfill_requests (tenant_id, scope_key, created_at DESC);
CREATE INDEX ix_backfill_running ON backfill_requests (tenant_id, source_id, job_type) WHERE status IN ('running', 'validating');
CREATE INDEX IF NOT EXISTS ix_backfill_status ON backfill_requests (tenant_id, status, created_at DESC);
CREATE INDEX IF NOT EXISTS ix_backfill_scope ON backfill_requests (tenant_id, scope_key, created_at DESC);
CREATE INDEX IF NOT EXISTS ix_backfill_running ON backfill_requests (tenant_id, source_id, job_type) WHERE status IN ('running', 'validating');
-- Processed Events: Duplicate suppression tracking (TTL-managed)
CREATE TABLE processed_events (
CREATE TABLE IF NOT EXISTS processed_events (
tenant_id TEXT NOT NULL,
scope_key TEXT NOT NULL,
event_key TEXT NOT NULL, -- Unique identifier for deduplication
@@ -97,14 +102,16 @@ CREATE TABLE processed_events (
CONSTRAINT pk_processed_events PRIMARY KEY (tenant_id, scope_key, event_key)
) PARTITION BY LIST (tenant_id);
CREATE TABLE processed_events_default PARTITION OF processed_events DEFAULT;
DO $$ BEGIN
CREATE TABLE processed_events_default PARTITION OF processed_events DEFAULT;
EXCEPTION WHEN duplicate_object OR SQLSTATE '42P17' THEN NULL; END $$;
CREATE INDEX ix_processed_events_expires ON processed_events (expires_at) WHERE expires_at < NOW() + INTERVAL '1 day';
CREATE INDEX ix_processed_events_time ON processed_events (tenant_id, scope_key, event_time DESC);
CREATE INDEX ix_processed_events_batch ON processed_events (tenant_id, batch_id) WHERE batch_id IS NOT NULL;
CREATE INDEX IF NOT EXISTS ix_processed_events_expires ON processed_events (expires_at);
CREATE INDEX IF NOT EXISTS ix_processed_events_time ON processed_events (tenant_id, scope_key, event_time DESC);
CREATE INDEX IF NOT EXISTS ix_processed_events_batch ON processed_events (tenant_id, batch_id) WHERE batch_id IS NOT NULL;
-- Backfill Checkpoints: Resumable batch processing state
CREATE TABLE backfill_checkpoints (
CREATE TABLE IF NOT EXISTS backfill_checkpoints (
checkpoint_id UUID NOT NULL,
tenant_id TEXT NOT NULL,
backfill_id UUID NOT NULL,
@@ -126,9 +133,11 @@ CREATE TABLE backfill_checkpoints (
CONSTRAINT ck_backfill_checkpoints_hash_hex CHECK (batch_hash IS NULL OR batch_hash ~ '^[0-9a-f]{64}$')
) PARTITION BY LIST (tenant_id);
CREATE TABLE backfill_checkpoints_default PARTITION OF backfill_checkpoints DEFAULT;
DO $$ BEGIN
CREATE TABLE backfill_checkpoints_default PARTITION OF backfill_checkpoints DEFAULT;
EXCEPTION WHEN duplicate_object OR SQLSTATE '42P17' THEN NULL; END $$;
CREATE INDEX ix_backfill_checkpoints_request ON backfill_checkpoints (tenant_id, backfill_id, batch_number);
CREATE INDEX IF NOT EXISTS ix_backfill_checkpoints_request ON backfill_checkpoints (tenant_id, backfill_id, batch_number);
-- Function to clean up expired processed events (called by background job)
CREATE OR REPLACE FUNCTION cleanup_expired_processed_events(batch_limit INTEGER DEFAULT 10000)
@@ -150,5 +159,3 @@ BEGIN
RETURN deleted_count;
END;
$$ LANGUAGE plpgsql;
COMMIT;