Lock performance

[mhenrixon]

While doing some profiling information on the locks, it became apparent that most of the locks' time was spent waiting for concurrent ruby, which was a bug in the wait time for a value.

Before:

After:

Hopefully, this will improve performance a bit as there should be a lot less waiting time.

[ViliusLuneckas]

Hi, anyone here experiencing randomly skipped jobs after this upgrade?

[ArturT]

@ViliusLuneckas Do you still see issues?

I tested it so far in development and it seemed to work fine. Didn't test on production yet.

[mhenrixon]

The number of skipped jobs should not increase per se but the effectiveness of the locking should be considerably increased. That said, it could be perceived as more jobs are skipped because there is less waiting time. That was a bug that cause high load systems to slow down.

The version with the performance optimization has been installed almost 26k times and I haven't had any complaints about it except this one comment.