background-jobs

Background jobs move slow work out of interactive request paths by using durable job contracts with defined state transitions, leases, and terminal outcomes. This ensures long-running work is processed safely without causing request timeouts.

Retries and idempotency in a job queue require idempotency keys to prevent duplicate execution and a retry policy with failure classification, exponential backoff, and jitter to safely recover from partial failures.

Checkpointing for resumable background workers saves job state machine progress periodically, allowing interrupted or failed tasks to resume from the last checkpoint rather than restarting, ensuring reliable state recovery.

User-visible progress and cancellation for serverless task offloading are supported by reporting meaningful progress signals like stages or counts, and defining clear status semantics for cancellation requests within the job contract.

A dead-letter queue for background job execution should be used when retries with exponential backoff are exhausted, capturing permanently failed jobs to prevent data loss and allow manual investigation or reprocessing.

Explicit job contracts define job identity, payload, state, attempts, progress, and leases to prevent invisible timeouts, while controlled concurrency limits parallel worker execution to avoid overwhelming downstream resources.