jteagles

@rbalamohan, could you take a look at this PR related to TezSplitGrouper when run on S3

@Darkness303, I have updated the PR title to better reach a reviewer. Can you confirm the title is correct and whether this occurs or S3 or another file system?

How often have you observed the ConcurrentModificationException? When I look at the code base, the callers of those functions are synchronized at the function level. Please 1) attach a stack...

@chenjunbiao001, can can provide a stack trace of the ConcurrentModificationException? I am unable to find a code path that will exercise the problem in practice.

@belugabehr, can you go back and run the performance tests in https://issues.apache.org/jira/browse/TEZ-1526. It will be interesting to see how this performs after removing the performance optimizations.

This code optimization was critically import as the the event thread spends a significant time parsing task/attempt ids to dispatch messages. I would hate to lose that. I can appreciate...

@belugabehr, In general, it could be a good change. Couple of thing on my mind about this. I have been thinking about replacing apache base64 due to performance after reading...

There is a Jenkins file in progress as a PR submission that I haven't had time to verify. If anyone has time to try this out, it may help with...

Something about this PR isn't correct as there are 58 commits. Are you targeting the correct branch?

Given the work around should this PR be closed in favor of TEZ-4074 ?