Re: Fwd: Stalled post to pgsql-es-ayuda - Mailing list pgsql-general

From Tom Lane
Subject Re: Fwd: Stalled post to pgsql-es-ayuda
Date
Msg-id 4478.1277354234@sss.pgh.pa.us
Whole thread Raw
In response to Fwd: Stalled post to pgsql-es-ayuda  (Sebastian Machuca <serroba@gmail.com>)
List pgsql-general
Sebastian Machuca <serroba@gmail.com> writes:
> The directory which contain the PGDATA/data is full, and i move the entire
> directory data to another partition, and i link symbolic to PGDATA.

> Now, when i try to start the server, keep on  "startup process   recovering
> "
> When i try to connect, this is the message:
> psql: FATAL:  the database system is starting up
> And never go on.

I think you just need to wait.  The thing that jumps out at me from your
details is

> Time of latest checkpoint:            Sun 30 May 2010 11:10:30 AM CLT

If it hasn't checkpointed since May, then either you've got some
seriously silly checkpoint settings or there was something blocking
checkpoints.  I'm guessing that it was the latter, and the reason you
ran out of disk space was that WAL files kept on accumulating (since
they can't be recycled until a checkpoint finishes).  So my theory is
that you have got a *whole lot* of WAL to churn through, and there is
nothing to do but wait for that to happen.  You could get an idea of how
long it will take by noting how fast the startup process is advancing
through WAL segments (watch its display in "ps") and looking to see what
the last WAL segment in $PGDATA/pg_xlog is.

Once you have a working database again, you should look into why
checkpoints were blocked.  Is there anything unusual in the postmaster
log from before you had the out-of-space condition?

            regards, tom lane

pgsql-general by date:

Previous
From: Konstantin Izmailov
Date:
Subject: Re: PQprepare in PostgreSQL 7.4 (lack of SAVEPOINTs)
Next
From: Scott Marlowe
Date:
Subject: Re: PQprepare in PostgreSQL 7.4 (lack of SAVEPOINTs)