Apache ZooKeeper server fails to form a quorum in Azure HDInsight

This article describes troubleshooting steps and possible resolutions for issues related to Zookeepers in Azure HDInsight clusters.

Symptoms

  • Both the resource managers go to standby mode
  • Namenodes are both in standby mode
  • Spark, Hive, and Yarn jobs or Hive queries fail because of Zookeeper connection failures
  • LLAP daemons fail to start on secure Spark or secure interactive Hive clusters

Sample log

You may see an error message similar to the following in yarn logs (/var/log/hadoop-yarn/yarn/yarn-yarn*.log on the headnodes):

2020-05-05 03:17:18.3916720|Lost contact with Zookeeper. Transitioning to standby in 10000 ms if connection is not reestablished.
Message
2020-05-05 03:17:07.7924490|Received RMFatalEvent of type STATE_STORE_FENCED, caused by org.apache.zookeeper.KeeperException$NoAuthException: KeeperErrorCode = NoAuth
...
2020-05-05 03:17:08.3890350|State store operation failed 
2020-05-05 03:17:08.3890350|Transitioning to standby state
  • High availability services like Yarn, NameNode, and Livy can go down for many reasons.
  • Confirm from the logs that it is related to Zookeeper connections
  • Make sure that the issue happens repeatedly (do not use these solutions for one off cases)
  • Jobs can fail temporarily due to Zookeeper connection issues

Common causes for Zookeeper failure

  • High CPU usage on the zookeeper servers
    • In the Ambari UI, if you see near 100% sustained CPU usage on the zookeeper servers, then the zookeeper sessions open during that time can expire and time out
  • Zookeeper clients are reporting frequent timeouts
    • In the logs for Resource Manager, Namenode and others, you will see frequent client connection timeouts
    • This could result in quorum loss, frequent failovers, and other issues

Check for zookeeper status

  • Find the zookeeper servers from the /etc/hosts file or from Ambari UI
  • Run the following command
    • echo stat | nc <ZOOKEEPER_HOST_IP> 2181 (or 2182)
    • Port 2181 is the apache zookeeper instance
    • Port 2182 is used by the HDInsight zookeeper (to provide HA for services that are not natively HA)
    • If the command shows no output, then it means that the zookeeper servers are not running
    • If the servers are running, the result will include statics of client connections and other statistics
Zookeeper version: 3.4.6-8--1, built on 12/05/2019 12:55 GMT
Clients:
 /10.2.0.57:50988[1](queued=0,recved=715,sent=715)
 /10.2.0.57:46632[1](queued=0,recved=138340,sent=138347)
 /10.2.0.34:14688[1](queued=0,recved=264653,sent=353420)
 /10.2.0.52:49680[1](queued=0,recved=134812,sent=134814)
 /10.2.0.57:50614[1](queued=0,recved=19812,sent=19812)
 /10.2.0.56:35034[1](queued=0,recved=2586,sent=2586)
 /10.2.0.52:63982[1](queued=0,recved=72215,sent=72217)
 /10.2.0.57:53024[1](queued=0,recved=19805,sent=19805)
 /10.2.0.57:45126[1](queued=0,recved=19621,sent=19621)
 /10.2.0.56:41270[1](queued=0,recved=1348743,sent=1348788)
 /10.2.0.53:59097[1](queued=0,recved=72215,sent=72217)
 /10.2.0.56:41088[1](queued=0,recved=788,sent=802)
 /10.2.0.34:10246[1](queued=0,recved=19575,sent=19575)
 /10.2.0.56:40944[1](queued=0,recved=717,sent=717)
 /10.2.0.57:45466[1](queued=0,recved=19861,sent=19861)
 /10.2.0.57:59634[0](queued=0,recved=1,sent=0)
 /10.2.0.34:14704[1](queued=0,recved=264622,sent=353355)
 /10.2.0.57:42244[1](queued=0,recved=49245,sent=49248)

Latency min/avg/max: 0/3/14865
Received: 238606078
Sent: 239139381
Connections: 18
Outstanding: 0
Zxid: 0x1004f99be
Mode: follower
Node count: 133212

CPU load peaks up every hour

  • Log in to the zookeeper server and check the /etc/crontab
  • If there are any hourly jobs running at this time, randomize the start time across different zookeeper servers.

Purging old snapshots

  • Zookeepers are configured to auto purge old snapshots
  • By default, the last 30 snapshots are retained
  • The number of snapshots that are retained, is controlled by the configuration key autopurge.snapRetainCount. This property can be found in the following files:
    • /etc/zookeeper/conf/zoo.cfg for Hadoop zookeeper
    • /etc/hdinsight-zookeeper/conf/zoo.cfg for HDInsight zookeeper
  • Set autopurge.snapRetainCount to a value of 3 and restart the zookeeper servers
    • Hadoop zookeeper config can be updated and the service can be restarted through Ambari
    • Stop and restart HDInsight zookeeper manually
      • sudo lsof -i :2182 will give you the process ID to kill
      • sudo python /opt/startup_scripts/startup_hdinsight_zookeeper.py
  • Do not purge snapshots manually - deleting snapshots manually could result in data loss

CancelledKeyException in the zookeeper server log doesn't require snapshot cleanup

  • This exception will be seen on the zookeeper servers (/var/log/zookeeper/zookeeper-zookeeper-* or /var/log/hdinsight-zookeeper/zookeeper* files)
  • This exception usually means that the client is no longer active and the server is unable to send a message
  • This exception also indicates that the zookeeper client is ending sessions prematurely
  • Look for the other symptoms outlined in this document

Next steps

If you didn't see your problem or are unable to solve your issue, visit one of the following channels for more support:

  • Get answers from Azure experts through Azure Community Support.
  • Connect with @AzureSupport - the official Microsoft Azure account for improving customer experience. Connecting the Azure community to the right resources: answers, support, and experts.
  • If you need more help, you can submit a support request from the Azure portal. Select Support from the menu bar or open the Help + support hub. For more detailed information, review How to create an Azure support request. Access to Subscription Management and billing support is included with your Microsoft Azure subscription, and Technical Support is provided through one of the Azure Support Plans.