waitinfuture opened a new pull request, #2621:
URL: https://github.com/apache/celeborn/pull/2621

   <!--
   Thanks for sending a pull request!  Here are some tips for you:
     - Make sure the PR title start w/ a JIRA ticket, e.g. '[CELEBORN-XXXX] 
Your PR title ...'.
     - Be sure to keep the PR description updated to reflect all changes.
     - Please write your PR title to summarize what this PR proposes.
     - If possible, provide a concise example to reproduce the issue for a 
faster review.
   -->
   
   ### What changes were proposed in this pull request?
   One of our users reported a dataloss issue in 
https://github.com/apache/celeborn/pull/2612 , I tried to reproduce
   the bug with the following setup:
   1. Partition data is far larger than 
`spark.celeborn.client.shuffle.partitionSplit.threshold`, which means split 
happens very often
   2. ``spark.celeborn.client.shuffle.partitionSplit.threshold` is larger than 
`celeborn.worker.shuffle.partitionSplit.max`, which means when split happens, 
it is `HARD_SPLIT`
   3. `celeborn.client.shuffle.batchHandleChangePartition.enabled` is true, 
which means when hard split happens, `LifecycleManager` will commit the splits 
before the stage finishes
   
   After digging into it, I found the bug is introduced by 
https://github.com/apache/celeborn/pull/2134 . #2134 added
   check in `InFlightRequestTracker#addBatch` and 
`InFlightRequestTracker#removeBatch` and only  
   increments/decrements `totalInflightReqs`  whether `batchIdSet` contains 
current `batchId`, which conflicts with
   `ShuffleClientImpl#PushDataRpcResponseCallback#updateLatestPartition`, which 
calls `addBatch` first then calls
   `removeBatch` with the same batchId. As a result, the call to `addBatch` 
fails to increment `totalInflightReqs`, but 
   the call to `removeBatch` decrements `totalInflightReqs`, which means the 
retried push is not counted, then later 
   `limitZeroInFlight` in `mapperEnd` will return even though the retried push 
fails.
   
   This PR fixes the bug by reverting #2134 
   
   
   ### Why are the changes needed?
   ditto
   
   
   ### Does this PR introduce _any_ user-facing change?
   No
   
   
   ### How was this patch tested?
   Manual test.
   


-- 
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