Mail Not Routing, stuck in /var/spool/mqueue.in

[Raymond Dijkxhoorn]

Hi!

> you've already send us the output of ps during a bad-performance event.
> This ps-output has stated somewhat clearly that you are suffering more
> or less sever disk IO problems:
>
> from 108 processes were 23 in "D" state. In case these processes are
> longer than a moment in "D" state this is (one part of) your problem.

Yes. But mainly caused by the 'loop'. The same ammount of mails on the
other box dont have this problem. I surely want to take for granted its
IO, but i really wonder why its not on the other box also, taking the
exact same ammount of mails and bytes. They are load balanced behind an
Alteon...

We also do log parsing, and when this is happening we also see MS scanning
more mails then come in (number of bytes) so MS is scanning some batches
over and over. Or perhaps i am missing something, but normally that
ammount should be pretty equal.

> In case these processes are just for short moment in "D" state, you've
> got much to much of them anyway. I don't know how to debug/improve your
> disk situation, but this seems to me of more need, than to debug MS ;-)

Could be, but i still think its not only io related. Sure, io shows up bad
when this is happening, but perhaps this is only the result, not the
cause. And yes, when its going on its like a snowball.

> Is all this "MS performs badly" just a Sobig-F story? Take 20% more
> Mails (and dubble Bytes) and the server went slow.

We block Sobig on the MTA allready, so no, its not Sobig, we reject
subjects...

Bye,
Raymond.