Re: [monit] Postgresql weirdness: ESTrootFATAL

[Martin Pala]

Hi,

pg_ctl needs to know the home of the database system, this can be provided to pg_ctl with the -D flag or through the environment variable $PGDATA.

As you mentioned, Monit sets spartan environment for start program (for security reasons).  If you started manually, the PGDATA was most probably set in your environment, whereas it's not that case in case of Monit. You can thus use the -D option for start and stop programs (pg_ctl)  in monit configuration.

Martin

On Nov 13, 2008, at 10:18 PM, David Paper wrote:

Hi monit gurus,

While I anxiously await 5.0 getting out of beta, I have run into the following problem w/ postgres: When started via monit, postgres spits out the following errors every 22 seconds in the postgres startup log::

2008-11-13 16:07:04 ESTrootFATAL:  database "root" does not exist
2008-11-13 16:07:04 ESTrootFATAL:  database "root" does not exist
2008-11-13 16:07:04 ESTrootFATAL:  database "root" does not exist
2008-11-13 16:07:04 ESTrootFATAL:  database "root" does not exist
2008-11-13 16:07:26 ESTrootFATAL:  database "root" does not exist
2008-11-13 16:07:26 ESTrootFATAL:  database "root" does not exist
2008-11-13 16:07:26 ESTrootFATAL:  database "root" does not exist
2008-11-13 16:07:26 ESTrootFATAL:  database "root" does not exist

4 entries in each group, forever.
Environment: Monit v 4.10.1 (started out of inittab), Postgres 8.3.4, SuSE Linux 11.0 x86-64.

This is what the monit job looks like:

check process postgresql with pidfile /opt/postgres/data/postmaster.pid
  group database
  start program = "/opt/postgres/bin/pg_ctl start"
as uid postgres and gid postgres
  stop  program = "/opt/postgres/bin/pg_ctl stop"
as uid postgres and gid postgres
  if failed unixsocket /tmp/.s.PGSQL.5432 protocol pgsql then restart
  if failed unixsocket /tmp/.s.PGSQL.5432 protocol pgsql then alert
  if failed host localhost port 5432             protocol pgsql then restart
  if failed host localhost port 5432             protocol pgsql then alert
  if 5 restarts within 5 cycles then timeout

I've also tried doing it like this:

check process postgresql with pidfile /opt/postgres/data/postmaster.pid
  group database
  start program = "/etc/init.d/postgresql start"
  stop  program = "/etc/init.d/postgresql stop"
  if failed unixsocket /tmp/.s.PGSQL.5432 protocol pgsql then restart
  if failed unixsocket /tmp/.s.PGSQL.5432 protocol pgsql then alert
  if failed host localhost port 5432             protocol pgsql then restart
  if failed host localhost port 5432             protocol pgsql then alert
  if 5 restarts within 5 cycles then timeout

Same result.

If I fire up postgres manually using the following methods, I don't see the error:

su - postgres; /opt/postgres/bin/pg_ctl start
su - postgres; /opt/postgres/bin/pg_ctl start -w -D /opt/postgres/data -l /opt/postgres/data/startup.log
su - postgres -c "LD_LIBRARY_PATH=/opt/postgres/lib /opt/postgres/bin/pg_ctl -w start -D \"/opt/postgres/data\" -l \"/opt/postgres/data/startup.log\""
(as root) /etc/init.d/postgresql start

It seems that only when fired up via a monit job is this an issue.

According to our DB guy, the error means that postgres is trying to find a database called "root", which of course, doesn't exist.

I know that Monit doesn't set any environment variables at time of start up of a jobs process. but I'm baffled as to where this is coming from.

Has anyone else that's running postgres seen this?

Thanks!

-dave

--
Dave Paper

"Hello, I must be going." --Groucho



--
To unsubscribe:
http://lists.nongnu.org/mailman/listinfo/monit-general