runaway processes, data corruption - Mailing list pgsql-admin

From Andy Marangakis
Subject runaway processes, data corruption
Date
Msg-id 417D4A96.2040109@colorado.edu
Whole thread Raw
List pgsql-admin
Hi all,
I've encounter a couple of bugaboos recently in a production database I run (7.3.4 on Solaris 8). One is that killing a query from a client tool such as dbVisualizer can cause the process on the server to become a runaway, consuming CPU and causing other processes to go into a TIME_WAIT state. The other is that a spontaneous reboot of the server (unknown cause, some hardware/OS wierdness) caused a catastrophic corruption, requiring rebuild from backup. The error returned was:

    PANIC: Invalid page header in block 2462 of 1365344463

I've upgraded to 7.4 in hopes that crash resistance improves. Any other thoughts on protecting against such occurances in the future? Thanks in advance.

_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/

Andy Marangakis                                   303-492-0527
Database Administrator                            303-492-4198 fax
Information Technology Services
University of Colorado at Boulder

pgsql-admin by date:

Previous
From: "Lee Wu"
Date:
Subject: Re: pg_ctl stop -m fast
Next
From: Shayne Patton
Date:
Subject: psql won't "createdb"