AbortTransaction - Mailing list pgsql-hackers

From Marcus Mascari
Subject AbortTransaction
Date
Msg-id 19981119200454.26289.rocketmail@send102.yahoomail.com
Whole thread Raw
List pgsql-hackers
I was hoping someone could shed some light on the 
following problem:

After increasing the buffer size of the postmaster
using the -B option to 256, we suffered a backend
crash which caused the backend to crash so hard,
no further connections could be made.

I rebooted the server (an intranet Web server-based
application), and attempted to reaccess the database
using the application.  The application had problems
querying one of the tables, where the query would 
simply suspend forever.  After rebooting again, I
decided to run a VACUUM ANALYZE manually (We run it
via cron every night), and received the following
message:

NOTICE:  AbortTransaction and not in in-progress state
NOTICE:  AbortTransaction and not in in-progress state

After perusing Usenet for similar problems, someone
had reported that they experienced the problem when
they increased their buffer size over 128K.  The
Usenet article was pre-6.4 (9/14/98).  I decreased the
buffer to 128, restarted the postmaster (I actually
rebooted), and reran VACUUM ANALYZE without problems.

Is it safe to increase the buffer size above 128K?
I did this because one of the queries (a 6-way join)
caused the backend to report an error telling me
to increase the buffer size.

Thanks for any info, 

Marcus Mascari (mascarim@yahoo.com)




_________________________________________________________
DO YOU YAHOO!?
Get your free @yahoo.com address at http://mail.yahoo.com



pgsql-hackers by date:

Previous
From: Terry Mackintosh
Date:
Subject: Search engine almost done, need spi knowlage
Next
From: Bruce Momjian
Date:
Subject: Re: [HACKERS] Bug in 6.4 release