Worker remains in Redis when JVM is killed about jesque HOT 9 CLOSED

See this - #11. Also we need dequeue method

from jesque.

I think, Jesque should use expire keys for workers and some cleaning up logic.
We can't be sure that application ends in safe way. And now there is no way to clean up wrong data in redis.

from jesque.

Does Resque have this feature? Do you have a suggestion as to how to make this cleanup happen safely?

from jesque.

Sorry. Resque... Jesque... =)

I'm talking about Jesque. As i understand, worker information is: key and values in some sets.
You can set expire for worker keys and use utility worker thread, that time to time updates expire information on worker keys that are started on current JVM, and cleans information from sets on workers, that are expired (keys not found).

from jesque.

So, if I understand you, your proposal is to have a separate thread on each worker JVM that updates an exipry key in Redis for each worker in that JVM. Then there is an external reaper process that periodically scans those exipry keys and if a key is stale, the reaper then assumes the corresponding worker's JVM has crashed hard and will cleanup all Redis keys associated with that worker. Is that correct?

Have you had an issue with your JVM's crashing? Have you tried adding a shutdown hook or a signal handler?

from jesque.

Have you had an issue with your JVM's crashing? Have you tried adding a shutdown hook or a signal handler?

Ofc, i use hooks. But they can't give you 100% guarantee =)

from jesque.

OK, just checking!
Would you be willing to start on this feature and submit a pull request?

from jesque.

@Kondratovich Any more thoughts on this?

from jesque.

@gresrun I am in the process of migrating from Resque to Jesque and have a heartbeat feature that needs to be migrated over - sounds like it fits this bill. Time permitting I'll try and get a PR for you.

from jesque.