xuanyuanking opened a new pull request #24110: [SPARK-25341][Core] Support 
rolling back a shuffle map stage and re-generate the shuffle files
URL: https://github.com/apache/spark/pull/24110
 
 
   ## What changes were proposed in this pull request?
   
   This is a follow-up work for #22112's future improvment[1]: `Currently we 
can't rollback and rerun a shuffle map stage, and just fail.`All changes are 
summarized as follows:
   - Extend ShuffleBlockId with indeterminateAttemptId.
   - Add corresponding support for ShuffleBlockResolver, if the shuffle file 
generated from the indeterminate stage, its name will contain the 
indeterminateAttemptId, otherwise the file name just as before.
   - Add the determinate flag in TaskContext and use it in Shuffle Reader and 
Writer.
   - Track the indeterminate attempt id in ShuffleStatus and add register and 
unregister support in MapOutputTracker
   - Add the determinate flag in Stage and use it in DAGScheduler. Also 
consider about the cleaning work for the intermediate state for the 
indeterminate stage.
   
   ## How was this patch tested?
   
   - UT: Add UT for all changing code and newly added function.
   - Manual Test:
   Also providing a manual test to verify the effect:
   ```
   import scala.sys.process._
   import org.apache.spark.TaskContext
   
   val determinateStage0 = sc.parallelize(0 until 1000 * 1000 * 100, 10)
   val indeterminateStage1 = determinateStage0.repartition(200)
   val indeterminateStage2 = indeterminateStage1.repartition(200)
   val indeterminateStage3 = indeterminateStage2.repartition(100)
   val indeterminateStage4 = indeterminateStage3.repartition(300)
   val fetchFailIndeterminateStage4 = indeterminateStage4.map { x =>
   if (TaskContext.get.attemptNumber == 0 && TaskContext.get.partitionId == 190 
&& 
     TaskContext.get.stageAttemptNumber == 0) {
     throw new Exception("pkill -f -n java".!!)
     }
     x
   }
   val indeterminateStage5 = fetchFailIndeterminateStage4.repartition(200)
   val finalStage6 = 
indeterminateStage5.repartition(100).collect().distinct.length
   ``` 
   It's a simple job with multi indeterminate stage, it will get a wrong anwser 
while useing old Spark version like 2.2/2.3, and will be killed after #22112. 
With this fix, the job can retry all indeterminate stage as below screenshot 
and get the right result.

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


With regards,
Apache Git Services

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

Reply via email to