Re: vacuums failing - Mailing list pgsql-general

From Tom Lane
Subject Re: vacuums failing
Date
Msg-id 1967.971810846@sss.pgh.pa.us
Whole thread Raw
In response to vacuums failing  (Joseph Shraibman <jks@selectacast.net>)
List pgsql-general
Joseph Shraibman <jks@selectacast.net> writes:
> Could someone please tell me what is going wrong?
> query: VACUUM  ANALYZE
> NOTICE:  FlushRelationBuffers(message, 15): block 0 is referenced
> (private 0, global 2)
> FATAL 1:  VACUUM (vc_repair_frag): FlushRelationBuffers returned -2

This is probably an artifact of the known bug that 7.0.* forgets to
run AbortTransaction() if a client disconnects in the middle of a
BEGIN block.  There's no harm to your data, but buffer reference counts
in shared memory aren't decremented before the backend exits.  So,
VACUUM gets unhappy because it fears some other backend is touching
the table it's vacuuming.

Restarting the postmaster should make the symptom go away.  The bug
itself is fixed for 7.0.3.

            regards, tom lane

pgsql-general by date:

Previous
From: "Martin A. Marques"
Date:
Subject: startup script for Solaris
Next
From: Tom Lane
Date:
Subject: Re: startup script for Solaris