[
https://issues.apache.org/jira/browse/NIFI-3574?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Bryan Rosander updated NIFI-3574:
---------------------------------
Description:
Currently if you point a PutHiveStreaming processor at a secure cluster, start
it, stop it, and then reconfigure it to point at a different unsecured cluster,
some state from the first connection is used by the underlying Hive and Hadoop
libraries when attempting to connect to the second cluster.
This can be mitigated by ensuring that the
UserGroupInformation.setConfiguration() method is used in onSetup() and that
the approprate UserGroupInformation instance is used in ugi.doAs calls.
In order to isolate multiple PutHiveStreaming processors, a
RequiresInstanceClassLoading annotation should be added to the processor. This
should insulate it from other static side effects.
was:
Currently if you point a PutHiveStreaming processor at a secure cluster, start
it, stop it, and then reconfigure it to point at a different unsecured cluster,
some state from the first connection is used by the underlying Hive and Hadoop
libraries when attempting to connect to the second cluster.
This can be mitigated by ensuring that the
UserGroupInformation.setConfiguration() method is used in onSetup() and that
the approprate UserGroupInformation instance is used in ugi.doAs calls.
In order to isolate multiple PutHiveStreaming processors, a
RequiresInstanceClassLoading should be added to the processor. This should
insulate it from other static side effects.
> PutHiveStreaming affected by UserGroupInformation static side effects
> ----------------------------------------------------------------------
>
> Key: NIFI-3574
> URL: https://issues.apache.org/jira/browse/NIFI-3574
> Project: Apache NiFi
> Issue Type: Bug
> Reporter: Bryan Rosander
> Assignee: Bryan Rosander
>
> Currently if you point a PutHiveStreaming processor at a secure cluster,
> start it, stop it, and then reconfigure it to point at a different unsecured
> cluster, some state from the first connection is used by the underlying Hive
> and Hadoop libraries when attempting to connect to the second cluster.
> This can be mitigated by ensuring that the
> UserGroupInformation.setConfiguration() method is used in onSetup() and that
> the approprate UserGroupInformation instance is used in ugi.doAs calls.
> In order to isolate multiple PutHiveStreaming processors, a
> RequiresInstanceClassLoading annotation should be added to the processor.
> This should insulate it from other static side effects.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
