Re: REINDEXing user table in non stand-alone? - Mailing list pgsql-admin

From Tom Lane
Subject Re: REINDEXing user table in non stand-alone?
Date
Msg-id 16722.1005229631@sss.pgh.pa.us
Whole thread Raw
In response to REINDEXing user table in non stand-alone?  ("Shuzo Kubo" <gardenweasel@hotmail.com>)
List pgsql-admin
"Shuzo Kubo" <gardenweasel@hotmail.com> writes:
> Does one need to go into stand-alone mode to
> reindex an user table?

No.  The docs need some improvement in this area.  The pushups mentioned
in the docs are really only necessary if you have a broken system index
that's sufficiently critical that the backend crashes before you can
execute REINDEX.  (For instance, the indexes on pg_class are likely to
cause such problems.)  The basic idea of the pushups is to start the
backend in a mode such that it doesn't use the indexes to access the
system catalogs.

> Is it safe to REINDEX an user table while in
> non stand-alone mode when other users are
> accessing it?

Not sure about that --- I believe it's safe in 7.2, but IIRC, REINDEX
didn't bother to acquire a LOCK on the table in earlier releases.
Best to shut down any other clients that might be trying to access the
damaged table.  Or forget REINDEX and recover via DROP/CREATE INDEX.

            regards, tom lane

pgsql-admin by date:

Previous
From: "Chad R. Larson"
Date:
Subject: Re: Problems with user-level security
Next
From: "Eric Sawler"
Date:
Subject: Logging to Syslog on RedHat 7.1