Re: Urgent Help required - Mailing list pgsql-general

From Jeff Davis
Subject Re: Urgent Help required
Date
Msg-id 1255726506.30077.172.camel@monkey-cat.sm.truviso.com
Whole thread Raw
In response to Urgent Help required  ("Neha Patel" <npatel@nevi-soft.com>)
Responses Re: Urgent Help required
List pgsql-general
On Fri, 2009-10-16 at 11:26 +0100, Neha Patel wrote:
> We are running with postgres sql 7.3.2. We were trying to create an
> index on a big table. The create index command ran for nearly 5 hours
> at which point we decided to interrupt it. Since this was interrupted,
> any operations attempted on the table on which the index was being
> created gives following error in pgsql log:

> LOG: all server processes terminated; reinitializing shared memory and
> semaphor

1. Shut down postgresql and make a full filesystem copy of the PGDATA
directory. This will ensure that anything else you do won't leave you in
a worse position.

2. Upgrade to the latest version of postgresql 7.3, which is 7.3.21

3. Start up again

4. Try to fix the problem:
 a. see if there are any indexes on the table
 b. if so, drop them
 c. try to get a good logical backup using pg_dump. You may want to
disable index scans by using "SET enable_indexscan = f;".
 d. If that doesn't work, you may have catalog corruption. Examine the
catalogs (documented here:
http://www.postgresql.org/docs/7.3/static/catalogs.html ), and look
entries related to your table in pg_class and pg_index, and see if
anything looks wrong.
 e. start from a fresh install and restore using the logical backup

5. Upgrade all data to a recent version of postgresql. You're on a
really ancient version that has fallen out of official support. This may
require several upgrade steps, but you should get upgraded to 8.3.8 or
8.4.1. A lot of bugs have been fixed, and as long as you are on 7.3, you
will still be at serious risk.

Regards,
    Jeff Davis


pgsql-general by date:

Previous
From: Raymond O'Donnell
Date:
Subject: Re: slightly off-topic: Central Auth
Next
From: Peter Hunsberger
Date:
Subject: Re: Partitioned tables as a poor mans columnar index?