cloudflare-queues

Cloudflare Queues enables asynchronous task processing on Workers by decoupling producers that push messages from consumers that process them. Create a queue, bind it to your Worker, send tasks from a producer, and deploy a consumer handler to process batches with built-in retry and error handling.

Cloudflare Queues supports exponential backoff retries and dead-letter queues (DLQs) for messages that exceed retry limits. Configure retry policies in your queue, route failures to a DLQ for inspection, and implement explicit acknowledgment to control when messages are marked complete.

Yes. Cloudflare Queues batches messages up to a configurable max batch size with timeout-based delivery, allowing you to group and process multiple tasks together. This reduces overhead and improves throughput for rate-limited APIs and large workloads.

Messages in Cloudflare Queues are limited to 128 KB each. Plan message size, batch window, and throughput accordingly; oversized payloads will be rejected and require payload restructuring or external storage references.

Use Wrangler CLI to define queues in wrangler.toml, bind them to Workers, and deploy producer and consumer code together. Wrangler handles queue provisioning, binding configuration, and local development.

Yes. Explicit acknowledgment lets you control when messages are marked complete, enabling robust error handling. Ack or retry messages based on processing results; unacknowledged messages are redelivered after the batch timeout.