Re: [Exim] Performance bottleneck scanning large spools.

Pàgina inicial
Delete this message
Reply to this message
Autor: Theo Schlossnagle
Data:  
A: exim-users
Assumpte: Re: [Exim] Performance bottleneck scanning large spools.
Thanks.. I guess I whould have posted the pertinent parts of my exim
configuration. I already have the split_spool_directory set to true.
We are running on ext2 and ext2 has notorious problems with large
directories. Unfortunamely, with 62 splits, the queue only needs to
grow by an order of magnitude to tickle the problems again. (When you
are dealing with 18 messages a second outgoing and spikes of up to 150
messages a second incoming, (relaying) you can see this happen too
often. I have on many occasions been stuck with a queue or over 2
million messages. I use different techiniques to solve thos problems.

Actually, I think we are running into some disk I/O problems as well..
I have 200-500 processes reading alot of information. They REALLY need
to be sharing what they read so that each disk block isn't read several
hundred times... Too many disk operations and too many system calls.

"Dave C." wrote:
>
> If you don't already, I'd try setting this option:
> | split_spool_directory
> |    Type:    boolean
> |    Default: false


--
Theo Schlossnagle
Senior Systems Engineer
33131B65/2047/71 F7 95 64 49 76 5D BA 3D 90 B9 9F BE 27 24 E7