[
https://issues.apache.org/jira/browse/HADOOP-18330?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17564599#comment-17564599
]
Steve Loughran commented on HADOOP-18330:
-----------------------------------------
looks like it was triggered by HADOOP-14556, though it's a bit more complex
than that as there's some canonicalisation taking place for delegation tokens
to work.
the way to fix this would be for an extra client attribute to be added to the
S3ClientCreationParameters and then passed in. that way, we can be sure there
will be no regressions.
patch welcome... i can see you are working in the code at a depth where you are
familiar with how to change this
> S3AFileSystem removes Path when calling createS3Client
> ------------------------------------------------------
>
> Key: HADOOP-18330
> URL: https://issues.apache.org/jira/browse/HADOOP-18330
> Project: Hadoop Common
> Issue Type: Bug
> Components: fs/s3
> Affects Versions: 3.3.0, 3.3.1, 3.3.2, 3.3.3
> Reporter: Ashutosh Pant
> Priority: Major
>
> when using hadoop and spark to read/write data from an s3 bucket like ->
> s3a://bucket/path and using a custom Credentials Provider, the path is
> removed from the s3a URI and the credentials provider fails because the full
> path is gone.
> In Spark 3.2,
> It was invoked as -> s3 = ReflectionUtils.newInstance(s3ClientFactoryClass,
> conf)
> .createS3Client(name, bucket, credentials);
> But In spark 3.3.3
> It is invoked as s3 = ReflectionUtils.newInstance(s3ClientFactoryClass,
> conf).createS3Client(getUri(), parameters);
> the getUri() removes the path from the s3a URI
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: common-issues-h...@hadoop.apache.org
