[
https://issues.apache.org/jira/browse/HADOOP-15735?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16609113#comment-16609113
]
Steve Loughran commented on HADOOP-15735:
-----------------------------------------
Seems reasonable. branch-2.7 is almost frozen apart from security patches (i.e.
spark should really move to 2.8+)
Even with a backport, you're going to have to rerun the whole s3a test suite.
Sorry, but as Yetus/jenkins doesn't we force all submitters (even ourselves) to
do their homework before others look at it
See:
https://hadoop.apache.org/docs/r2.7.6/hadoop-aws/tools/hadoop-aws/index.html
(don't worry about the s3n/s3 tests, just leave their configs blank and ignore
the errors from those tests)
> backport HADOOP-11687 for hadoop-aws 2.7.x
> ------------------------------------------
>
> Key: HADOOP-15735
> URL: https://issues.apache.org/jira/browse/HADOOP-15735
> Project: Hadoop Common
> Issue Type: Bug
> Components: fs/s3
> Affects Versions: 2.7.7
> Reporter: MunShik JOUNG
> Priority: Trivial
> Attachments:
> 0001-HADOOP-12970-Intermitent-SignatureDoesNotMatch-Excep.patch
>
>
> HADOOP-12970 Intermitent SignatureDoesNotMatch Exception is fixed by
> HADOOP-11687.
> And from hadoop-aws >= 2.8, HADOOP-11687 is applied. but Spark 2.3.1 is still
> using hadoop-aws 2.7.7.
> I was stumpled upon Intermitent SignatureDoesNotMatch Exception with s3 with
> spark 2.3.1. So I guess backporting might benefit some spark users. In my
> situation, SignatureDoesNotMatch was reliably reproducible (4 out of 8
> attemp). And after using patched version symptom was not seen (0 out of 10)
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
