HyukjinKwon opened a new pull request, #58:
URL: https://github.com/apache/spark-connect-rust/pull/58

   ### What changes were proposed in this pull request?
   
   Start the CI Spark Connect server with
   `spark.connect.execute.reattachable.senderMaxStreamDuration=1200s` in the
   Python-Connect-parity workflow.
   
   ### Why are the changes needed?
   
   After #54 merged, `test_parity_arrow_python_udf.py` failed the master parity 
gate with
   a whole-file **TIMEOUT (p=0)** — zero tests completed in the 1200s per-file 
cap — even
   though the exact same file passes **deterministically locally in ~113s (320 
passed, 64
   skipped)**, matching the reference client's ~120s. So it is not a client 
correctness bug
   and not mere slowness (it ran the full 1200s doing nothing).
   
   Root cause: on the loaded single-node CI server, a slow first 
arrow-Python-UDF worker
   lets the default `senderMaxStreamDuration` elapse, so the server ends the 
reattachable
   `ExecutePlan` response stream early and the client must `ReattachExecute`; 
that reattach
   path stalls under these conditions and hangs the whole file. Raising the 
sender duration
   keeps the stream open until results are produced, so no reattach is 
triggered and the
   file completes normally. (The underlying reattach-path robustness is worth 
hardening in
   the client separately; this unblocks the gate without masking a client bug.)
   
   ### Does this PR introduce _any_ user-facing change?
   
   No — CI configuration only.
   
   ### How was this patch tested?
   
   CI: the Python Connect parity job on this PR. Locally the file already 
passes in ~113s.
   
   This pull request and its description were written by Isaac.
   


-- 
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]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to