Dead-lettering carries over the TTL into the new queue

[rjrudman]

If you set an expiration on a message, and the queue has x-dead-letter-exchange or x-dead-letter-routing-key set, when the message is pushed to the new queue, the TTL still persists.

This means that if you set up a queue to watch for dead-lettered items, if it's not processed quickly enough these messages will vanish forever.

Perhaps there should be a parameter to set the new TTL after dead-lettering, or remove the TTL after this occurs altogether?

[hyperthunk]

Hmn, this is an interesting question. I suspect this won't happen if you set TTL on the queue rather than the individual messages, but even for per-message-ttl you raise an interesting question about how the semantics should work. We will have a discussion about this and let you know which way we're going shortly.

Thanks for reporting it!

[hyperthunk]

In conclusion, we've decided to remove the 'expiration' property from messages that are dead-lettered, since we consider this less surprising behaviour from a user perspective. The fix is visible in http://hg.rabbitmq.com/rabbitmq-server/rev/ecd2b82b69cc and will be included in the next release.

Thanks again for the report!

[rjrudman]

Thank you very much for the response and quick turn around! The fix looks great

[hyperthunk]

Fix is due to go out in 3.1.2, which should come out in the next couple of days.

[hyperthunk]

This was included in today's release - see the release notes for details. Thanks again for raising it.