From our experience it is not just a postgres issue, but all IO with the Update 3 kernel.
We have a box with Update 3 that queries a remote postgres database(Running RH7.3, RH3 Update2) and writes to a file on an NFS server. The update 3 box does half the work with 2-3 times the load as our update 1 and 2 boxes. Looking at top the box is always above 90% IO Wait on the CPU. When we downgrade the kernel to Update 2 it seems to fix the issue. We several Update 3 boxes that run postgres locally and they all struggle compared to the Update 2 boxes We have tried the Fedora Core 3 with not much more success and we are going to try the Update 4 beta kernel next week to see if it is any better. There are several threads on the Taroon mailing list discussing the issue. Woody -----Original Message----- From: Josh Berkus [mailto:[EMAIL PROTECTED] Sent: Thursday, November 18, 2004 1:34 PM To: [EMAIL PROTECTED] Cc: Woody Woodring; 'Andrew Janian' Subject: Re: [PERFORM] Query Performance and IOWait Woody, > What version of Redhat are you running? We have found running > Enterprise Update 3 kernel kills our Dell boxes with IOWait, both NFS > and local disk traffic. Update 2 kernel does not seem to have the > issue, and we are in the process of trying Update 4 beta to see if it is better. This is interesting; do you have more to say about it? I've been having some mysterious issues with RHES that I've not been able to pin down. -- Josh Berkus Aglio Database Solutions San Francisco ---------------------------(end of broadcast)--------------------------- TIP 9: the planner will ignore your desire to choose an index scan if your joining column's datatypes do not match
