What information is required to file a Jira for Greenplum RCA request for failed segments or standby master?

posted Apr 28, 2017, 4:58 PM by Sachchida Ojha
Failed segments or standby master requires root cause analysis (RCA)

If the customer requests for an RCA for failed segments or standby master on the Greenplum database, the following information must be collected and a Jira must be created with L3 collaboration.

Running gpdetective is the most preferred way, but in case there are issues running it, the following is the required information:
  • The actual problem details
  • Which version of Greenplum the customer is running
  • Output of gp_segment_configuration in case of 4.x or gp_configuration in case of 3.x
  • Output of gp_configuration_history table.
  • Master logs with the error happened
  • Segment logs of the failed segments incase of segment failures (Standby master logs inc ase of failed standby)
  • Segment log files of the corresponding primaries or mirrors of the failed segments. This information can be got fromgp_segments_configuration table where content id is the same as that of failed segments.
  • Core files if any created
  • dmesg- from the failed segments
  • iostat -nE - failed segments
  • netstat -i
  • /var/log/messages
  • If this a Solaris system - zpool list 
  • df -h - from the failed segments

  • The results of gpcheckperf if the customer is willing to take a down time.