Retry_Logic.md

Grading jobs retry system

This module provides a DB-backed retry mechanism for failed grading jobs.

How it works

• A grading_jobs table stores failed job payloads and metadata (attempts, nextRetryAt, etc.).
• GradingJobService polls the table periodically and replays jobs by calling GradingResultService.saveResult.
• Exponential backoff is used between attempts. After maxAttempts the job is marked FAILED.

Configuration & running

• No external dependencies (Redis) required — the system uses the primary database (TypeORM).
• The module is auto-registered in AppModule and will start polling when the application boots.

Usage

• To enqueue a failed grading job from other code, inject GradingJobService and call enqueueFailedJob(submissionId, payload, maxAttempts?).

Testing

• The service exposes processOnce() for tests to trigger a single retry pass.

Notes

• Ensure autoLoadEntities or explicit entity registration is enabled so TypeORM picks up GradingJobEntity.
• Consider adding retention/cleanup for old failed jobs and hooks for alerting on repeated failures.