eugenegujing commented on PR #6724:
URL: https://github.com/apache/texera/pull/6724#issuecomment-5249685042

   @Yicong-Huang Hi Yicong, I think we need a second look on the CI read: the 
`ReconfigurationIntegrationSpec` failures appear to be caused by this PR. The 
job failed on both runners on every push of this branch across three main 
bases, while main/merge-queue runs are green (42/42), and I reproduced it 
locally with tracing. 
   
   The tests pause before the slow source's first tuple, so the Fries 
reconfiguration ECM is the first-ever message on the source→udf data channel, 
which now registers born-disabled and swallows the marker. `updateExecutor` is 
never acked and `reconfigureWorkflow` times out at 30s (on main this only works 
*because of* the leak this PR fixes). 
   
   The `PAUSED→COMPLETED` tracebacks are secondary fallout from two 
pre-existing potential bugs (`_check_and_process_control` processes ECMs under 
a stale `current_input_channel_id`, and `PauseManager.resume` re-enables 
ECM_PAUSE channels before checking `_global_pauses`). So this PR might need 
more work. I'd either move the leak fix to the dequeue side (stash a leaked 
DataElement, retro-disable its channel, re-inject on resume) or settle the 
"markers depend on the leak" question at the design level first. What do you 
think?


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to