Re: [SQL] Can VACUUM, but cannot do VACUUM ANALYZE - Mailing list pgsql-sql

From Tom Lane
Subject Re: [SQL] Can VACUUM, but cannot do VACUUM ANALYZE
Date
Msg-id 1391.940695513@sss.pgh.pa.us
Whole thread Raw
In response to Can VACUUM, but cannot do VACUUM ANALYZE  ("Albert REINER" <areiner@tph.tuwien.ac.at>)
List pgsql-sql
"Albert REINER" <areiner@tph.tuwien.ac.at> writes:
>> albert=> vacuum analyze;
>> NOTICE:  AbortTransaction and not in in-progress state
>> pqReadData() -- backend closed the channel unexpectedly.

This unhelpful trace is the result of the interaction of a couple
of different bugs --- first, vacuum is encountering some sort of
error; then error recovery is messing up in the context of vacuum
(leading first to the NOTICE and then to the coredump); and when
libpq observes connection closure, it drops the original error
message (which it has already received!) on the floor in its haste
to tell you about the closed connection.

The error recovery problem is fixed as of 6.5.2, I believe, so it'd
be worth your while to upgrade.  The libpq misbehavior is fixed for
7.0, but it's part of some extensive changes that seemed too risky
to back-patch into 6.5.*.  In the meantime, the only way to find out
what that original error message was is to consult the postmaster log
file --- you are running the postmaster with stdout/stderr going to
some logfile, I hope.  It should show up right before the NOTICE and
then the postmaster's response to the backend coredump.

Once we have vacuum's error message we can maybe offer some useful
counsel.
        regards, tom lane


pgsql-sql by date:

Previous
From: "Oliver Elphick"
Date:
Subject: Re: [SQL] Can VACUUM, but cannot do VACUUM ANALYZE
Next
From: Tom Lane
Date:
Subject: Re: [SQL] sum-function