Re: [exim] Spamassassin + Exim4 high IO Wait

Kezdőlap
Üzenet törlése
Válasz az üzenetre
Szerző: Mark Adams
Dátum:  
Címzett: Jeremy Harris
CC: exim-users
Tárgy: Re: [exim] Spamassassin + Exim4 high IO Wait
On Tue, Sep 19, 2006 at 01:25:53PM +0100, Jeremy Harris wrote:
> Mark Adams wrote:
> > I was hoping to draw from the pool of experience that the exim mailing
> > list has, as I realise this is a spamassassin or other issue. We are
> > running exim4.62-4 with Spamassassin 3.1.4-4 and suffering extremely
> > high IO Wait times on the server whenever spamassassin scans an incoming
> > email. Exim4 is setup as defaults (spamd running locally on standard
> > port) so no configuration was changed. Spamassassin has not had any
> > rules added to it. The server goes from 1-2% wait times spiking to
> > between 20-90% when spamd is scanning an email, this lasts for a few
> > seconds and then drops back down to normal. This is a system with raided
> > SATA drives on a LVG that also serves small SMB shares. It serves around
> > 40 mail users.
> >
> > I have not had this problem on servers with SCSI drives, or on servers
> > that use SATA but are strictly mail servers.
>
> Traditionally I/O wait time is, as far as the processor is concerned,
> just another sort of idle time. If you had more cpu work to do they'd
> be doing it.


I'm talking about the wait time on the disks (iowait) which as defined
by iostat manpage

%iowait
"Show the percentage of time that the CPU or CPUs were idle during which
the system had an outstanding disk I/O request."

it seems the disks cannot keep up with spamassassin?

It is definatly something that I am worrying about as it is causing lag
to the SMB service on the same server.

>
> As to why you've not seen it with SCSI drives, I'd suspect buggy
> drivers not reporting waits properly.
>
> In short - you're doing I/O. Don't worry.
>
> - Jeremy
>
> --
> ## List details at http://www.exim.org/mailman/listinfo/exim-users
> ## Exim details at http://www.exim.org/
> ## Please use the Wiki with this list - http://www.exim.org/eximwiki/