What you'll have at the end

A dead letter queue holding the original job and a full record of every failed attempt

Before you start

You need

A task that already retries itself a fixed number of times before giving up, running behind a queue you already have working.

Not covered

Automatically replaying a dead-lettered job once its cause is fixed; this stops at making sure the failure and its full history are captured and someone is told.

Leans on

Background Job Processing Systems Cheat Sheetthe Dead letter queue row lives in its Retry and Failure Handling Strategies table, and DLQ size sits in Job Queue Monitoring and Observability, both tables this recipe builds out directly.
Message Queues Cheat Sheetthe Message Deduplication row covers what has to stay true before a dead-lettered job is ever sent again, background for reading the record in the closing steps.

Where to go next

Catch the one poison message that retries forever and starves every worker

for a job that never counts down its retries at all and just keeps hammering forever, not one that quietly gives up after a fixed number.

Exponential backoff stops a retry storm from hammering a flaky API

for spacing the retries themselves out so they don't hammer a struggling API in the same second, a concern that sits before this recipe's own retries ever run out.

Give a poison message its own dead-letter queue

for the same idea at the raw broker level, routing individual messages instead of framework-managed jobs.

A missing dead letter queue means an exhausted job just vanishes

Before you start

Where to go next

A missing dead letter queue means an exhausted job just vanishes

Before you start

Where to go next