David Throckmorton <throckwork@gmail.com> writes:
> Following a recent database crash, I encountered the following when doing
> debug:
> [root@my-db myhost]# ps -ef | grep pgsql
> postgres 19643 5737 0 09:42 ? 00:00:00
> /usr/pgsql-9.0/bin/postmaster -p 5432 -D /var/lib/pgsql/9.0/data
> postgres 19644 5737 0 09:42 ? 00:00:00
> /usr/pgsql-9.0/bin/postmaster -p 5432 -D /var/lib/pgsql/9.0/data
> postgres 19657 5737 0 09:42 ? 00:00:00
> /usr/pgsql-9.0/bin/postmaster -p 5432 -D /var/lib/pgsql/9.0/data
> postgres 19658 5737 0 09:42 ? 00:00:00
> /usr/pgsql-9.0/bin/postmaster -p 5432 -D /var/lib/pgsql/9.0/data
> pgstartup.log doesn't indicate anything about multiple instances starting,
> and indicates that PID 5737 was the initial PID for the instance:
Almost certainly, these are just regular postmaster child processes.
I'm not sure why "ps" is showing you the original postmaster command
line rather than the usual modified process title --- that's a known
behavior on some Unixen but I've not heard of it on Linux. Are you
using a nonstandard flavor of "ps"?
Also, is process 5737 still around? If it was dead, I would rather
expect ps to show the orphaned processes as children of pid 1.
If it is around, what is ps showing as its command line? Can you
connect to the database?
regards, tom lane