Pivotal Knowledge Base

Follow

HAWQ Cluster Start Fails giving Error "sh: /home/gpadmin/hawq-data-directory/masterdd/pg_log/startup.log: No such file or directory"

Environment

Product Version
 Pivotal HDB  2.x

Symptom

HAWQ start fails with a "No such file or directory" error:

$ hawq start cluster
20161021:11:47:50:099428 hawq_start:hawq20dn2:gpadmin-[INFO]:-Prepare to do 'hawq start'
20161021:11:47:50:099428 hawq_start:hawq20dn2:gpadmin-[INFO]:-You can find log in:
20161021:11:47:50:099428 hawq_start:hawq20dn2:gpadmin-[INFO]:-/home/gpadmin/hawqAdminLogs/hawq_start_20161021.log
20161021:11:47:50:099428 hawq_start:hawq20dn2:gpadmin-[INFO]:-GPHOME is set to:
20161021:11:47:50:099428 hawq_start:hawq20dn2:gpadmin-[INFO]:-/usr/local/hawq/.
20161021:11:47:50:099428 hawq_start:hawq20dn2:gpadmin-[INFO]:-Start hawq with args: ['start', 'cluster']
20161021:11:47:50:099428 hawq_start:hawq20dn2:gpadmin-[INFO]:-Gathering information and validating the environment...
20161021:11:47:50:099428 hawq_start:hawq20dn2:gpadmin-[INFO]:-No standby host configured
20161021:11:47:50:099428 hawq_start:hawq20dn2:gpadmin-[INFO]:-Start all the nodes in hawq cluster
20161021:11:47:50:099428 hawq_start:hawq20dn2:gpadmin-[INFO]:-Starting master node 'localhost'
20161021:11:47:50:099428 hawq_start:hawq20dn2:gpadmin-[INFO]:-Start master service
20161021:11:52:51:099428 hawq_start:hawq20dn2:gpadmin-[INFO]:-sh: /home/gpadmin/hawq-data-directory/masterdd/pg_log/startup.log: No such file or directory
pg_ctl: PID file "/home/gpadmin/hawq-data-directory/masterdd/postmaster.pid" does not exist
20161021:11:52:52:099428 hawq_start:hawq20dn2:gpadmin-[ERROR]:-Master start failed, exit

Cause

The /usr/local/hawq/etc/hawq-site.xml contains the path to a HAWQ master/ segment directory that does not exist. 

This can happen after an upgrade if the upgrade was completed, and either:

  • The /usr/local/hawq/etc/ directory was not recovered from the previous installation - if the HAWQ cluster is NOT managed via Ambari.

    or

  • The HAWQ cluster IS managed via Ambari, but the cluster is being restarted via command line.


Resolution

IF THE CLUSTER IS MANAGED VIA AMBARI: 

  • In Ambari under HAWQ / Config, check the values for the configuration settings "HAWQ Segment Directory" and "HAWQ Master Directory"

  • Check on all segments and masters to make sure that the directories exist and that the permissions are correct:
# ls -ld /data/hawq/master/
drwx------. 16 gpadmin gpadmin 4096 Oct 21 12:00 /data/hawq/master/

  • Start the cluster via Ambari instead of the command line.

 

IF THE CLUSTER IS MANAGED VIA COMMAND LINE, AND THE CLUSTER HAS BEEN UPGRADED:

  • Copy the configuration from the previous version of HAWQ:
hawq_node$ cp -rf /usr/local/hawq-2.0.0.0/etc /usr/local/hawq/
hawq_node$ chown -R gpadmin:gpadmin /usr/local/hawq/etc
  • Try to restart HAWQ via command line again:
hawq_node$ hawq start cluster

 

 

Comments

Powered by Zendesk