Maxence MaireauxandGitHub bc0e5976c2 feat: scalable retry mechanism with atomic claim pattern (#141) ...

Redesign the retry worker to support multi-worker scaling with continuous
batch processing instead of all-at-once polling.

Key changes:
- Add "retrying" intermediate status for atomic claim pattern via CTE
- Batch processing with configurable LIMIT (--retry-batch-size, default 50)
- Reduce tick interval from 1min to 3s (--retry-period) for continuous flow
- Parallel batch processing via pond worker pool (bounded concurrency)
- Crash recovery for stale "retrying" attempts (runs once per minute)
- Partial indexes for retry polling, claimed attempts, and recovery
- UpdateAttemptsStatus scoped to "retrying" rows only (preserves history)
- Worker resilient to per-webhook errors (log + continue, no crash)
- HTTP client timeout (30s) prevents blocking on slow endpoints
- Oldest retries prioritized via ORDER BY next_retry_after ASC

2026-04-01 12:09:48 +02:00

..

module.go

feat: scalable retry mechanism with atomic claim pattern (#141)

2026-04-01 12:09:48 +02:00