[
https://issues.apache.org/jira/browse/AMBARI-3966?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Yusaku Sako updated AMBARI-3966:
--------------------------------
Summary: Secure cluster: set group owner for the task-controller /
container-executor binary baesd on appropriate settings in mapred-site.xml /
yarn-site.xml (was: Secure cluster: task-controller binary should be owned by
the group specified in taskcontroller.cfg in Hadoop 1 and yarn)
> Secure cluster: set group owner for the task-controller / container-executor
> binary baesd on appropriate settings in mapred-site.xml / yarn-site.xml
> ----------------------------------------------------------------------------------------------------------------------------------------------------
>
> Key: AMBARI-3966
> URL: https://issues.apache.org/jira/browse/AMBARI-3966
> Project: Ambari
> Issue Type: Bug
> Components: agent
> Affects Versions: 1.4.2
> Reporter: Jaimin D Jetly
> Assignee: Jaimin D Jetly
> Fix For: 1.4.2
>
> Attachments: AMBARI-3966.patch, AMBARI-3966_2.patch
>
>
> In order to set the group for the task-controller binary in Hadoop 1 or
> container-executor binary in Hadoop 2:
> h5. For Hadoop 1.x stack secure cluster:
> * Stop MapReduce service.
> * Reconfigure mapreduce.tasktracker.group property in MapReduce > Configs >
> Advanced section.
> * Manually create the group per the value entered in
> mapreduce.tasktracker.group on each tasktracker host.
> * Add mapred user to that group.
> * Start the MapReduce service.
> * This reconfigures the mapreduce.tasktracker.group property in
> /etc/hadoop/taskcontroller.cfg and also sets the group ownership for the
> /usr/lib/hadoop/bin/task-controller binary.
> * Run MapReduce smoke test. Smoke test should pass.
> h5. For Hadoop 2.x stack secure cluster:
> * Stop Yarn service.
> * Reconfigure yarn.nodemanager.linux-container-executor.group property in
> YARN > Configs > Advanced section.
> * Manually create the group per the value entered in
> yarn.nodemanager.linux-container-executor.group on each nodemanager host.
> * Add yarn user to that group.
> * Start the YARN service.
> * This reconfigures the yarn.nodemanager.linux-container-executor.group
> property in /etc/hadoop/container-executor.cfg and also set the group value
> for the /usr/lib/hadoop-yarn/bin/container-executor binary.
> * Run YARN smoke test. Smoke test should pass.
--
This message was sent by Atlassian JIRA
(v6.1#6144)
