wuchang created YARN-6590:
-----------------------------
Summary: ResourceManager Master/Slave transition make all
applications killed
Key: YARN-6590
URL: https://issues.apache.org/jira/browse/YARN-6590
Project: Hadoop YARN
Issue Type: Bug
Affects Versions: 2.7.3
Environment: Linux
Reporter: wuchang
Priority: Critical
My yarn is configured as HA . It seems that because of the zk connection
timeout , the active ResourceManager become standby and the standby one become
active,namely , the ResourceManager active/standby transition. But both the
process of two RM is OK . Below is the ResourceManager error log :
{noformat}
2017-05-12 12:47:40,150 INFO
org.apache.hadoop.yarn.server.resourcemanager.security.NMTokenSecretManagerInRM:
Sending NMToken for nodeId : 10.120.117.100:37900 for container :
container_1494505293131_4378_01_000007
2017-05-12 12:47:40,150 INFO
org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainerImpl:
container_1494505293131_4378_01_000007 Container Transitioned from ALLOCATED to
ACQUIRED
2017-05-12 12:47:40,150 INFO
org.apache.hadoop.yarn.server.resourcemanager.security.NMTokenSecretManagerInRM:
Sending NMToken for nodeId : 10.120.117.108:46066 for container :
container_1494505293131_4378_01_000008
2017-05-12 12:47:40,150 INFO
org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainerImpl:
container_1494505293131_4378_01_000008 Container Transitioned from ALLOCATED to
ACQUIRED
2017-05-12 12:47:40,166 INFO org.apache.zookeeper.ClientCnxn: Opening socket
connection to server 10.120.117.104/10.120.117.104:2181. Will not attempt to
authenticate using SASL (unknown error)
2017-05-12 12:47:40,168 INFO org.apache.zookeeper.ClientCnxn: Socket connection
established to 10.120.117.104/10.120.117.104:2181, initiating session
2017-05-12 12:47:40,170 INFO org.apache.hadoop.ha.ActiveStandbyElector: Session
expired. Entering neutral mode and rejoining...
2017-05-12 12:47:40,170 INFO org.apache.zookeeper.ClientCnxn: Unable to
reconnect to ZooKeeper service, session 0x685bcd9343dfc3f8 has expired, closing
socket connection
2017-05-12 12:47:40,170 INFO org.apache.hadoop.ha.ActiveStandbyElector: Trying
to re-establish ZK session
{noformat}
In my opinion , this active/standby transition *should not* make my running
application killed , but in fact , when this transition happened , all the
running YARN-BASED MR and Spark jobs are killed. Below is some of my yarn
configuration.
{code}
<property>
<name>yarn.resourcemanager.zk-address</name>
<value>zkServer1:2181,zkServer2:2181,zkServer3:2181,zkServer4:2181</value>
</property>
<property>
<name>yarn.resourcemanager.zk-timeout-ms</name>
<value>30000</value>
</property>
<property>
<name>yarn.resourcemanager.store.class</name>
<value>org.apache.hadoop.yarn.server.resourcemanager.recovery.ZKRMStateStore</value>
</property>
<property>
<name>yarn.resourcemanager.ha.automatic-failover.enabled</name>
<value>true</value>
</property>
{code}
So , any configuration missing?I notice that I didn't configure the
{noformat}yarn.resourcemanager.recovery.enabled{noformat} to true and the
default value is false.But according to the official document , this
configuration is used for ResourceManager restart, instead of for
ResourceManager Active/Standby transition.
Any suggestions?
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
