[
https://issues.apache.org/jira/browse/HADOOP-16163?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16783508#comment-16783508
]
Steve Loughran commented on HADOOP-16163:
-----------------------------------------
Looking at this, I'd point the blame clearly at whatever the test is doing
regarding kerberos minicluster setup and reset.
Why do we do that? Mainly because the ABFS client will only serve up DTs if
kerberos is enabled.
It doesn't have to do that. HDFS does it because it is relying on Kerberos for
auth. If ABFS has a DT provider which can serve up OAuth tokens, then it can do
this even while Kerberos is off, and it radically simplifies testing, as
there's no need to do things with kerberos miniclusters.
Note: MR and Spark will still only collect DTs if kerberos is enabled, but they
are free to change that behaviour.
Is there a risk of problems if we disable DT issuing? Only in the case of: A DT
provider was declared that is only expected to run if kerberos is enabled. Any
provider like that would have to check for kerberos running and downgrade
-though really you'd want the cluster setup to be consistent here.
> NPE in setup/teardown of ITestAbfsDelegationTokens
> --------------------------------------------------
>
> Key: HADOOP-16163
> URL: https://issues.apache.org/jira/browse/HADOOP-16163
> Project: Hadoop Common
> Issue Type: Sub-task
> Components: fs/azure, test
> Affects Versions: 3.3.0
> Reporter: Steve Loughran
> Assignee: Steve Loughran
> Priority: Major
>
> Managed to create this in a local (variant) branch:
> {Code}
> [ERROR] Errors:
> [ERROR]
> org.apache.hadoop.fs.azurebfs.extensions.ITestAbfsDelegationTokens.testCanonicalization(org.apache.hadoop.fs.azurebfs.extensions.ITestAbfsDelegationTokens)
> [ERROR] Run 1: ITestAbfsDelegationTokens.setup:107 » NullPointer
> [ERROR] Run 2: ITestAbfsDelegationTokens.teardown:130 » NullPointer
> {code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
