On Wed, 2009-04-15 at 09:51 -0400, Tom Lane wrote:
> Brian Cox <brian.cox@ca.com> writes:
> > I changed the logic to update the table in 1M row batches. However,
> > after 159M rows, I get:
>
> > ERROR: could not extend relation 1663/16385/19505: wrote only 4096 of
> > 8192 bytes at block 7621407
>
> You're out of disk space.
>
> > A df run on this machine shows plenty of space:
>
> Per-user quota restriction, perhaps?
>
> I'm also wondering about temporary files, although I suppose 100G worth
> of temp files is a bit much for this query. But you need to watch df
> while the query is happening, rather than suppose that an after-the-fact
> reading means anything.
Anytime we get an out of space error we will be in the same situation.
When we get this error, we should
* summary of current temp file usage
* df (if possible on OS)
Otherwise we'll always be wondering what caused the error.
--
Simon Riggs www.2ndQuadrant.com
PostgreSQL Training, Services and Support