# Scheduler Graph Jobs: PostgreSQL Migration Plan (2025-12-06)

## Goals
- Replace Mongo-based GraphJobStore/PolicyRunService with PostgreSQL equivalents.
- Keep graph job determinism (status transitions, ordering) and tenant isolation.
- Provide schema, repository surface, and migration steps to unblock PG-T7.1.2a (Cleanup Wave A).

## Proposed Schema (schema: `scheduler`)
- `graph_jobs`
  - `id UUID PK`
  - `tenant_id TEXT NOT NULL`
  - `type SMALLINT NOT NULL` (0=build,1=overlay)
  - `status SMALLINT NOT NULL` (queued, running, completed, failed, canceled)
  - `payload JSONB NOT NULL` (serialized GraphBuildJob/GraphOverlayJob)
  - `created_at TIMESTAMPTZ NOT NULL DEFAULT now()`
  - `updated_at TIMESTAMPTZ NOT NULL DEFAULT now()`
  - `correlation_id TEXT NULL`
  - Indexes: `idx_graph_jobs_tenant_status` (tenant_id, status, created_at DESC), `idx_graph_jobs_tenant_type_status` (tenant_id, type, status, created_at DESC)

- `graph_job_events`
  - `id BIGSERIAL PK`
  - `job_id UUID NOT NULL REFERENCES graph_jobs(id) ON DELETE CASCADE`
  - `tenant_id TEXT NOT NULL`
  - `status SMALLINT NOT NULL`
  - `payload JSONB NOT NULL`
  - `created_at TIMESTAMPTZ NOT NULL DEFAULT now()`
  - Index: `idx_graph_job_events_job` (job_id, created_at DESC)

## Repository Contracts
- `IGraphJobRepository` (Postgres)
  - `ValueTask InsertAsync(GraphBuildJob job, CancellationToken ct)`
  - `ValueTask InsertAsync(GraphOverlayJob job, CancellationToken ct)`
  - `ValueTask<bool> TryReplaceAsync(GraphBuildJob job, GraphJobStatus expected, CancellationToken ct)`
  - `ValueTask<bool> TryReplaceOverlayAsync(GraphOverlayJob job, GraphJobStatus expected, CancellationToken ct)`
  - `ValueTask<GraphBuildJob?> GetBuildJobAsync(string tenantId, string id, CancellationToken ct)`
  - `ValueTask<GraphOverlayJob?> GetOverlayJobAsync(string tenantId, string id, CancellationToken ct)`
  - `ValueTask<IReadOnlyCollection<GraphBuildJob>> ListBuildJobsAsync(string tenantId, GraphJobStatus? status, int limit, CancellationToken ct)`
  - `ValueTask<IReadOnlyCollection<GraphOverlayJob>> ListOverlayJobsAsync(string tenantId, GraphJobStatus? status, int limit, CancellationToken ct)`
  - `ValueTask AppendEventAsync(GraphJobEvent evt, CancellationToken ct)`

## Migration
- New migration file: `014_graph_jobs.sql` under `src/Scheduler/__Libraries/StellaOps.Scheduler.Storage.Postgres/Migrations` with the tables above.

## DI Changes
- Replace `AddSchedulerMongoStorage` and `MongoGraphJobStore` in WebService with `AddSchedulerPostgresStorage` and new `PostgresGraphJobStore` implementing `IGraphJobStore`.
- Worker.Backfill: swap Mongo options to Postgres options; use Postgres repos from `StellaOps.Scheduler.Storage.Postgres`.

## Tests
- Add Postgres integration tests for `PostgresGraphJobRepository` covering insert/list/update/expected-status checks and event log.
- Update WebService/Worker tests to use Postgres fixtures; remove Mongo fixtures.

## Rollback
- If regressions occur, revert migration + DI switch; Mongo storage remains in history.

## Owners
- Schema/repo: Scheduler Guild
- DI/tests: Scheduler Guild