Upon running gpstart, the start process fails with an unknown error.

Post date: Oct 05, 2012 3:11:48 PM

A gpstop was executed, but gpsyncmaster is still running on smdw.

20110105:18:33:46:gpstart:mdw:gpadmin-[INFO]:-Starting standby master 20110105:18:33:46:gpstart:mdw:gpadmin-[INFO]:-Checking if standby master is running on host: smdw in direc 20110105:18:33:46:gpstart:mdw:gpadmin-[INFO]:-No db instance process, entering recovery startup mode 20110105:18:33:47:gpstart:mdw:gpadmin-[INFO]:-Commencing parallel primary and mirror segment instance start 20110105:18:36:55:gpstart:mdw:gpadmin-[INFO]:-Process results... 20110105:18:36:55:gpstart:mdw:gpadmin-[CRITICAL]:-gpstart failed. (Reason='') exiting...

“gpstart -m” You can run this command when there is no master postgres process running on the master host. This is the process that is listening on port 5432 (default). It will fail if a postgres process is already running on that port. This gives you the ability to read the Greenplum system tables, and we would take advantage of the gp_segment_configuration table to run a parallel instruction. “gpstop -af” This means that a parallel stop of all Greenplum segments will be started on both segment hosts and the standby master. Other forms of gpstop will leave thegpsyncmaster process running on the standby master host. This is a known bug when there are complications with the gpsyncmasterprocess. The stop includes “f” the force option, which will terminate postgres process and drop client connections if they exist. When this command completes all processes, including the problematic gpsyncmaster processes, should be stopped on segment hosts.

“gpstart -a” This command will bring the database back up, successfully this time since no gpsyncmaster process id running on standby. The “-a” in these commands mean “do not prompt.”