[YARN-11512] Graceful decommission doesn't work when NM restart recovery is enabled

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 3.3.1
• Component/s: graceful, nodemanager
• Assignee: Unassigned
• Reporter: Akshesh Doshi
• Created: 13/Jun/23 20:35
• Updated: 13/Jun/23 21:02

Description

We have added these configs on yarn-site.xml file of our Hadoop-Yarn cluster.

<property>
    <name>yarn.nodemanager.recovery.enabled</name>
    <value>true</value>
</property>
<property>
    <name>yarn.nodemanager.recovery.supervised</name>
    <value>true</value>
</property>

The NM restart recovery feature has been working well, applications not failing even if we restart nodemanager processes. But, when we try to decommission a node by adding the node name to yarn_exclude_hosts file and refreshing nodes on resourcemanager, the applications that had containers running on that node are stuck for a long time and then fail.