Re: No control over max.num. WAL files

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 05/25/2011 02:55 PM, Craig Ringer wrote:
>> These are the relevant parameters we have changed in postgresql.conf:
>>
>> archive_mode | off
>> checkpoint_segments | 128
>> default_statistics_target | 100
>> maintenance_work_mem | 512MB
>> max_fsm_pages | 800000
>> max_fsm_relations | 8000
>> shared_buffers | 10GB
>> wal_buffers | 512kB
>> wal_sync_method | fdatasync
>> work_mem | 16MB
>
>
> Are you sure you don't also have WAL archiving enabled? That can easily
> lead to too many WAL segments being kept if the archiving process can't
> keep up or fails.
>

I am sure. I have double checked.

# grep archive_mode postgresql.conf
archive_mode = off

# SHOW archive_mode;
archive_mode
- --------------
off
(1 row)

> It'd really help if you could show part of your postgresql logs from the
> during restore.
>

This happens when restoring a 30GB database. It looks like it stops when
creating a GIN index on a 4217 MB table. See my first post for the
relevant information from the log file (No relevant information/errors
before this)

STATEMENT: CREATE INDEX attachments_textsearch ON attachments USING gin
(textsearchable);

Where 'textsearchable' is a tsvector column. In the source database,
this index is 7989 MB.

regards,
- --
Rafael Martinez Guerrero
Center for Information Technology
University of Oslo, Norway

PGP Public Key: http://folk.uio.no/rafael/
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.14 (GNU/Linux)

iEYEARECAAYFAk3dAIIACgkQBhuKQurGihSRUQCfTHrPPoc0ean2O99269KOIEZY
fjIAoIHoVIw4QOx0s52wWy4XBh9gH1Os
=N0TH
-----END PGP SIGNATURE-----