How to collect information on hung Greenplum Database processes

posted Apr 28, 2017, 4:48 PM by Sachchida Ojha
Collect this information:
ps ax output for all postgres processes on master/all nodes:

gpssh -f <hosts_file> "hostname; ps ax | grep postgres"


pstack/gstack output for all postgres processes on master/all nodes:

gpssh -f <hosts_file> "ps ax | grep postgres | awk '{ print \$1}' \| while read pid; do hostname; echo $pid; gstack \$pid; done"


df -h/zfs list output for all postgres processes on master/all nodes:

gpssh -f <hosts_file> "df -h"


pg_locks from master and segments:

select * from pg_locks;
select * from gp_distrandom('pg_locks');

pg_stat_activity from master and segments:

select * from pg_stat_activity;
select * from gp_distrandom('pg_stat_activity');
Comments