If the job dies, it doesn't remove the lock

[MaicolBen]

Describe the bug
This happens with until_executed because the job needs to be executed

Expected behavior
The lock should be released

Current behavior
The lock persists forever

Additional context
Gem version: 6.0.0.rc7

[mhenrixon]

That is expected behaviour. You mean the job was retried any number of times configured like retry and not completing? I'll add info in readme how to deal with that.

[mhenrixon]

See this section @MaicolBen https://github.com/mhenrixon/sidekiq-unique-jobs/blob/master/README.md#cleanup-dead-locks

[MaicolBen]

I can use sidekiq_retries_exhausted, but if I adduntil_executed lock, it doesn't enter to sidekiq_retries_exhausted code after moving it to the dead queue. This doesn't happen with until_executing

[MaicolBen]

Well after some time, I got that the callbacks calls with retry: 0, the documentation is outdated, the method is del instead of delete_by. But it seems the lock is somewhere because isn't fired again:

SidekiqUniqueJobs::Digests.all(pattern: '3606e62e8d702dd382c873f64a543d40') # => []
MyWorker.perform_async(id) # => nil

[mhenrixon]

Ah yes! I put the wrong method in the documentation. Will update it in the next PR!

[MaicolBen]

I did it on #311, do you know why is still locked if the digest doesn't exist anymore?

[mhenrixon]

So I was looking into this @MaicolBen. It seems Redis is just not very reliable with the seconds. In my tests a sleep outside the map block seems to work fine but it is unreliable. You probably want something like 1 minute or so for the lock_expiration to make it more robust and reliable.