[AMBARI-21204] Yarn stopped by itself after start. HA run

• Type: Bug
• Status: Resolved
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 2.5.1
• Fix Version/s: 2.5.2
• Assignee: Dmytro Sen
• Reporter: Dmytro Sen
• Created: 08/Jun/17 16:36
• Updated: 12/Jun/17 16:05
• Resolved: 12/Jun/17 15:22

Description

From RM logs :

2017-06-07 14:23:19,191 FATAL resourcemanager.ResourceManager (ResourceManager.java:main(1240)) - Error starting ResourceManager
org.apache.hadoop.service.ServiceStateException: java.io.IOException: Couldn't set ACLs on parent ZNode: /yarn-leader-election
        at org.apache.hadoop.service.ServiceStateException.convert(ServiceStateException.java:59)
        at org.apache.hadoop.service.AbstractService.init(AbstractService.java:172)
        at org.apache.hadoop.service.CompositeService.serviceInit(CompositeService.java:107)
        at org.apache.hadoop.yarn.server.resourcemanager.AdminService.serviceInit(AdminService.java:152)
        at org.apache.hadoop.service.AbstractService.init(AbstractService.java:163)
        at org.apache.hadoop.service.CompositeService.serviceInit(CompositeService.java:107)
        at org.apache.hadoop.yarn.server.resourcemanager.ResourceManager.serviceInit(ResourceManager.java:281)
        at org.apache.hadoop.service.AbstractService.init(AbstractService.java:163)
        at org.apache.hadoop.yarn.server.resourcemanager.ResourceManager.main(ResourceManager.java:1236)
Caused by: java.io.IOException: Couldn't set ACLs on parent ZNode: /yarn-leader-election
        at org.apache.hadoop.ha.ActiveStandbyElector.ensureParentZNode(ActiveStandbyElector.java:351)
        at org.apache.hadoop.yarn.server.resourcemanager.EmbeddedElectorService.serviceInit(EmbeddedElectorService.java:103)
        at org.apache.hadoop.service.AbstractService.init(AbstractService.java:163)
        ... 7 more
Caused by: org.apache.zookeeper.KeeperException$BadVersionException: KeeperErrorCode = BadVersion for /yarn-leader-election

The problem is that disabling security changes zk ACL for resource manager as part of AMBARI-19331. After the recent change in HDFS-11403, RM checks znode version and fails if it's different than expected.
The correct fix could be to remove znode during security disabling and do not break election znode consistency by manually changing ACL to all. RM should create it with proper ACL.

Issue Links

• links to: Review Board