alanjds / celery-serverless Goto Github PK

The _client_invoke_watchdog() call checks for the client lock and then for the watchdog lock, before truly invoking the watchdog.

A possible optimization is to let the client lock acquired for a bit more of time after a success watchdog invocation, as the next clients trying to acquire/check the locks should fail anyway. Why let the 1st lock open for someone just to see that the 2nd lock is closed?

Points of discussion:

• How much time should it be kept locked?
• How to unlock after the time passed? Automatically via timeout?
• How to deal with locks that does not support automatic unlock on timeout?

Update README so that other people can use this package. Can someone guide me on how to use this package?

Is the 2nd time that the watchdog refuses to work because a lock is forgotten on the Redis lock backend. It should have an automatic unlock timeout. (EXPIRE key?)