Re: Less available diskspace after crashed CLUSTER

"Sander Steffann" <sander(at)steffann(dot)nl> writes:
> 2005-01-15 15:17:46 [30605] PANIC: PANIC: could not write to file
> "/var/lib/pgsql/data/pg_xlog/xlogtemp.30605": Geen ruimte over op apparaat
> STATEMENT: CLUSTER;

Hmm. I wonder why we have XLogFileInit forcing a PANIC for this. At
one time it was called only from critical sections and so the error
would become a panic anyway, but that's not true anymore...

> The database is running nice again, but I have the feeling I lost some
> diskspace... I suspected a temp file, so I stopped the postmaster, looked in
> /var/lib/pgsql/data/base/17142/pgsql_tmp, found one file and deleted it. But
> now I still have 1G less available diskspace than before the CLUSTER.

> Where should I look next? I would like the diskspace back :-)

I'm guessing that the new table/index files created during the CLUSTER
didn't get released after the PANIC. Look for files that are not
referenced by any relfilenode value in pg_class. Don't forget that
there may be multiple sections (filenode.1, etc). See
http://developer.postgresql.org/docs/postgres/storage.html
(which is for 8.0, but everything except the material on tablespaces
applies to 7.4).

Also, did you get rid of the xlog temp file mentioned in the message?

regards, tom lane