swamped mail relays - suggestions

[Steen, Glenn]

Steen, Glenn wrote:
> Greg Matthews wrote:
>> On Tue, 2005-05-17 at 16:04 +0200, Steen, Glenn wrote:
>>> Greg Matthews wrote:
>>> (snipped mangled quote by OutLook-QuoteFix. Sigh)
>> 
>> as bad as groupwise?
> 
> Been so since last I looked at GW that I really couldn't tell:-).
> But it mangled the iostat beyond recognition.
> 
>> 
>>>> Any comments and suggestions most welcome.
>> 
>>> Seems that write performance might be what is killing you.
>>> Do you use the tmpfs thing for the incoming (MS work-) directory?
>> 
>> yes, I already tmpfs for the work dir. I am thinking that with 10,000
>> mails in the mqueue.in directory (~20000 files) the system might be
>> stalling just reading the dir.
>> 
>> I've just increased max children from 9 to 20 to see if that helps.
>> The comments recommend 5 per cpu but doesnt mention HT processors.
>> 
>>> 
>>> -- Glenn
> 
> Ok, that's good then.
> And you might well be right that 2x10K-file-directories might be the
> problem (Not perhaps for MS as much as the actual MTA)... Have you
> turned off atime updating too?
> 
> -- Glenn

... And perhaps try getting rid of ext3 journal updating too (as
suggested by David), by "reverting" to ext2.

-- Glenn

------------------------ MailScanner list ------------------------
To unsubscribe, email jiscmail at jiscmail.ac.uk with the words:
'leave mailscanner' in the body of the email.
Before posting, read the Wiki (http://wiki.mailscanner.info/) and
the archives (http://www.jiscmail.ac.uk/lists/mailscanner.html).

Support MailScanner development - buy the book off the website!