Implement retry mechanism for failed analysis jobs

[coderabbitai]

Problem

The database schema includes retry fields (retry_count, max_retries) but the worker doesn't implement retry logic:

• Worker immediately marks jobs as FAILED without checking retry count
• No exponential backoff between retries
• BullMQ built-in retry not configured
• TIMEOUT state in schema is unused

Action Items

1. Worker-level retries:

   • Check retry_count < max_retries before marking FAILED
   • Increment retry_count and re-enqueue job with backoff
   • Only mark FAILED when max retries exceeded

2. BullMQ configuration:

   • Configure job attempts in queue options
   • Add backoff strategy (exponential)
   • Set job timeout to use TIMEOUT state

3. Error classification:

   • Distinguish transient (network) vs permanent (invalid data) errors
   • Only retry transient errors
   • Fail fast for permanent errors

Example Implementation

const analysisWorker = new Worker(
  'analysis',
  async (job) => { /* ... */ },
  {
    connection: createBullMQConnection(),
    settings: {
      backoffStrategy: (attemptsMade) => Math.min(1000 * 2 ** attemptsMade, 60000),
    }
  }
);

// In job options when enqueuing:
await analysisQueue.add('analyze', data, {
  attempts: 3,
  timeout: 300000, // 5 minutes
  backoff: { type: 'exponential', delay: 2000 }
});

Files

• backend/worker/analysis.worker.js
• backend/controllers/webhook/handleWebhook.js
• backend/prisma/schema.prisma

Related

• PR Idempotent pr analysis built #28 (deferred retry to v3)
• Comment: Idempotent pr analysis built #28 (comment)

Requested by: @yb175