fix: pipeline pausing race conditions of draining and terminating source

[kohlisid]

Explain what this PR does.

[juliev0]

I'm curious - are you able to explain now what could've caused pl.Status.Phase to be set to "Paused" here but not reflected in etcd Pipeline Status?

[kohlisid]

From the logs I captured, I was able to see the status getting reflected to Paused in memory for sure. In the next reconciliation process (which begins almost immediately) that got triggered we started from running again and due to a race we stay in pausing as we are not able to drain out the buffers.

[juliev0]

From the logs I captured, I was able to see the status getting reflected to Paused in memory for sure. In the next reconciliation process (which begins almost immediately) that got triggered we started from running again and due to a race we stay in pausing as we are not able to drain out the buffers.

That's interesting but I still don't understand what would cause the write to etcd not to occur (unless it is writing and the watch is skipping changes).

[codecov]

Codecov Report

Attention: Patch coverage is 27.77778% with 26 lines in your changes missing coverage. Please review.

Project coverage is 64.06%. Comparing base (d340a4e) to head (b8b55bf).
Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
pkg/reconciler/pipeline/controller.go	27.77%	24 Missing and 2 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2131      +/-   ##
==========================================
- Coverage   64.10%   64.06%   -0.04%     
==========================================
  Files         325      325              
  Lines       31558    31574      +16     
==========================================
- Hits        20229    20227       -2     
- Misses      10277    10295      +18     
  Partials     1052     1052              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[juliev0]

LGTM as far as I can tell

[whynowy]

The logic change looks good to me, it would be better to have more appropriate variable and function names.