[
https://issues.apache.org/jira/browse/HADOOP-16207?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16944488#comment-16944488
]
Steve Loughran commented on HADOOP-16207:
-----------------------------------------
Final patch replaces the committer-specific terasort and MR test jobs with
parameterization
of the (now single tests) and use of file:// over hdfs:// as the cluster FS.
The parameterization ensures that only one of the specific committer tests
run at a time -overloads of the test machines are less likely, and so the
suites can be pulled back into the parallel phase.
There's also more detailed validation of the stage outputs of the terasorting;
if one test fails the rest are all skipped. This and the fact that job
output is stored under target/yarn-${timestamp} means failures should
be more debuggable.
We also have the s3guard operations log enabled; on guarded runs this tracks
all PUT/DELETE/TOMBSTONE calls made of a store, so acts as the log of what
changes were made there. If we see intermittent issues here again (And after
the HADOOP-16570 changes) then we are better positioned to understand the
failures
> Improved S3A MR tests
> ---------------------
>
> Key: HADOOP-16207
> URL: https://issues.apache.org/jira/browse/HADOOP-16207
> Project: Hadoop Common
> Issue Type: Sub-task
> Components: fs/s3, test
> Affects Versions: 3.3.0
> Reporter: Steve Loughran
> Assignee: Steve Loughran
> Priority: Critical
>
> Reported failure of {{ITestDirectoryCommitMRJob}} in validation runs of
> HADOOP-16186; assertIsDirectory with s3guard enabled and a parallel test run:
> Path "is recorded as deleted by S3Guard"
> {code}
> waitForConsistency();
> assertIsDirectory(outputPath) /* here */
> {code}
> The file is there but there's a tombstone. Possibilities
> * some race condition with another test
> * tombstones aren't timing out
> * committers aren't creating that base dir in a way which cleans up S3Guard's
> tombstones.
> Remember: we do have to delete that dest dir before the committer runs unless
> overwrite==true, so at the start of the run there will be a tombstone. It
> should be overwritten by a success.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: common-issues-h...@hadoop.apache.org
