"I/O error on connection" problem. MailScanner related?

[Chris W. Parker]

Julian Field <mailto:MailScanner at ecs.soton.ac.uk>
    on Wednesday, July 19, 2006 11:42 AM said:

> The "uptime" load averages (also printed by "top") are a vague
> indication of system load, but don't worry if these are much greater
> than 1. If they are less than 1 then your system definitely isn't
> loaded. Check your sendmail settings in /etc/mail/sendmail.cf. Look
> for QueueLA and other load averages limits (which all contain LA in
> their name). It's usually worth increasing them as MailScanner can
> push the load average up to 15 quite easily when fully loaded and
> working nicely. So don't start refusing messages until the load
> average is really quite high.

Thanks Julian. I'm going to experiment with this for a few days and see
what happens. As an extreme test I set both QueueLA and RefuseLA to 0.

> The virus scanners should only use CPU very briefly, they usually
> aren't significant at all in the load of MailScanner. Try switching
> off the biggies such as SpamAssassin and see how it speeds up.

As I don't want to lose spam detection or virus detection I also tried
changing the Queue Scan Interval from 10 seconds to 30. I don't recall
what the default value was (perhaps 10?) but I imagine that at some
point it's more efficient to check the queue often rather than let the
queue build up quite large and check it only periodically. Is there any
rule of thumb to this or does it vary too greatly from system to system?



Chris.