[Dovecot] Tools to analyze status/stats?

Chris Wakelin c.d.wakelin at reading.ac.uk
Wed Dec 7 00:00:56 EET 2005


Timo Sirainen wrote:
> On Tue, 2005-12-06 at 08:02 -0500, Stewart Dean wrote:
> 
>>Much of my aggravations with user use/misuse/stupidity traces back to 
>>this simple fact: there is rarely any feedback mechanism in server apps 
>>to indicate to the user the loads their practices cause.  It is as if 
>>they were driving a car without a gas gauge, if they had bodies that 
>>didn't feel pain.
>>/I/ wish that there was some metric that could tell a user the composite 
>>CPU and memory loads their mail usage represents.
> 
> 
> Hmm. I'm not sure if such loads can be told to use in any easy way. But
> I've been thinking about having a plugin that could tell admin what's
> happening in the computer. What imap/pop3 processes are using all the
> disk I/O, memory and CPU and what exactly they're doing, and if there
> are some special conditions (eg. broken indexes causing the load, etc).
> 

I've been using pacct/acctcom etc. on Solaris to give per-process CPU
usage, disk block count, and kcore minutes. It's a bit fiddly though!

Talking of which, I've been trying IMAP SSL using both Perdition as a
proxy to a Dovecot without SSL support, and Dovecot directly with SSL
support and, to my surprise, found Dovecot-SSL using much more CPU for
longer-lived processes (as much in an afternoon as my username usually
uses in a week!). The Perdition proxy method seems to add a 30-50%
overhead to the CPU load. I need to get some more users to test it all,
though!

Best Wishes,
Chris

-- 
--+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+-
Christopher Wakelin,                           c.d.wakelin at reading.ac.uk
IT Services Centre, The University of Reading,  Tel: +44 (0)118 378 8439
Whiteknights, Reading, RG6 2AF, UK              Fax: +44 (0)118 975 3094


More information about the dovecot mailing list