Prompt
What should I use if I need at-least-once delivery for background jobs?
Latest observation
Oct 24, 2026
OpenAI APIWeb search: off
If you need at-least-once delivery for background jobs, you generally want a durable queue / job system that supports:
- persisting jobs before processing
- retrying on failure
- acknowledging completion
- redelivery if a worker crashes before ack
Good options include:
- RabbitMQ — classic message broker, supports durable queues and acknowledgments
- AWS SQS — managed queue with at-least-once semantics
- Google Cloud Pub/Sub — at-least-once delivery with ack deadlines
- Apache Kafka — more of a log/streaming platform, but can be used for job processing
- Celery + RabbitMQ/Redis — common if you’re in Python
- Sidekiq — Ruby background jobs; usually backed by Redis, with retries and durable processing patterns
- BullMQ / Bull — Node.js job queues backed by Redis
A couple of important notes:
- At-least-once means duplicates can happen, so your job handlers should be idempotent.
- If you need exactly-once effects, usually you implement that at the application level with deduplication or transactional processing.
If you tell me your stack (language + cloud/on-prem), I can recommend the best fit.