Yi Zhao wrote:
> I'm sorry for the lack of information given in this mail;
> Postgresql: 8.3.3
> System:Linux 2.6.9-55.ELsmp
> Install: I compile it myself
>
> Others:
> the same command, it works fine on another machine(called A):(
>
> I EXPLAIN the sql machine A and machine B, I found that on machine B, it
> never use index when command execute, but on machine A, index used:((,
> so I "set enable_seqscan TO off;" on machine B, it's works beacause of
> the usage of index!!!
>
> I don't konw why.
It almost certainly means that you have a corrupt index. First, stop the
server and make a full backup of your data directory.
Then start the server back up. Do a full pg_dump if you can.
Now, try issuing a REINDEX on the problem index. If you don't know which
one, or want to make sure, use REINDEX DATABASE instead. Re-test and see
if the problem still occurs.
You should also check your RAID controller, disks, and filesystem to
make sure there's no problem in your storage system. If you are not
using a RAID controller with battery backup cache, make sure write
caching is turned off on the controller and the disks.
If you are not using a UPS, consider getting one.
By the way, a Google search (while I was trying to figure out which
distro and version of the distro you were using) turned this up:
http://www.centos.org/modules/newbb/viewtopic.php?topic_id=8779&forum=27
The crash is in ext3, and was triggered by MySQL. See:
http://bugs.centos.org/view.php?id=2077
So: check your system logs for errors from the kernel. Consider
upgrading to a less ancient kernel, too.
--
Craig Ringer