[ https://issues.apache.org/jira/browse/BEAM-7864?focusedWorklogId=302652&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-302652 ]
ASF GitHub Bot logged work on BEAM-7864: ---------------------------------------- Author: ASF GitHub Bot Created on: 28/Aug/19 08:16 Start Date: 28/Aug/19 08:16 Worklog Time Spent: 10m Work Description: iemejia commented on issue #9410: [BEAM-7864] Simplify/generalize Spark reshuffle translation URL: https://github.com/apache/beam/pull/9410#issuecomment-525635634 Thanks Kyle. @RyanSkraba raises some valid points. There is something weird in our current translation and the fact that we ignore keys in particular for the `Reshuffle.viaRandomKey()` case. We should maybe fill a JIRA to track this + discuss in the mailing list. (some [previous discussion on Reshuffle here](https://lists.apache.org/thread.html/820064a81c86a6d44f21f0d6c68ea3f46cec151e5e1a0b52eeed3fbf@%3Cdev.beam.apache.org%3E)). I was also wondering to what extent in our current implementation (and in particular for the random key case) we could do a repartition with more partitions (based on available CPUs). Of course this has the risk of eating more resources than defined by the job but on the other hand it could be a way to optimize such shuffles downstream. [but well this is a different subject just thinking] ---------------------------------------------------------------- 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: us...@infra.apache.org Issue Time Tracking ------------------- Worklog Id: (was: 302652) Time Spent: 2h 40m (was: 2.5h) > Portable Spark Reshuffle coder cast exception > --------------------------------------------- > > Key: BEAM-7864 > URL: https://issues.apache.org/jira/browse/BEAM-7864 > Project: Beam > Issue Type: Bug > Components: runner-spark > Reporter: Kyle Weaver > Assignee: Kyle Weaver > Priority: Major > Labels: portability-spark > Fix For: 2.16.0 > > Time Spent: 2h 40m > Remaining Estimate: 0h > > running :sdks:python:test-suites:portable:py35:portableWordCountBatch in > either loopback or docker mode on master fails with exception: > > java.lang.ClassCastException: org.apache.beam.sdk.coders.LengthPrefixCoder > cannot be cast to org.apache.beam.sdk.coders.KvCoder > at > org.apache.beam.runners.spark.translation.SparkBatchPortablePipelineTranslator.translateReshuffle(SparkBatchPortablePipelineTranslator.java:400) > at > org.apache.beam.runners.spark.translation.SparkBatchPortablePipelineTranslator.translate(SparkBatchPortablePipelineTranslator.java:147) > at > org.apache.beam.runners.spark.SparkPipelineRunner.lambda$run$1(SparkPipelineRunner.java:96) > at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511) > at java.util.concurrent.FutureTask.run(FutureTask.java:266) > at > java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149) > at > java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624) > at java.lang.Thread.run(Thread.java:748) -- This message was sent by Atlassian Jira (v8.3.2#803003)