On Sun, May 09, 2010, Steve Shockley wrote: > A few days ago, I had an old Windows box that worked as an inbound > mail relay start to fail, so I figured I'd replace it with two > OpenBSD boxes in a CARP pool.
Oops... usually you replace 10 windows boxes with a single Unix server... > The site gets about 30-40k messages per day. During periods of > heavy load, the load average would occasionally spike over 12, and Hmm, this isn't Linux, so disk I/O isn't counted, right? Anyway, the default is probably not a good value. Moreover, LA isn't really useful (as you know), e.g., I run a multi-threaded MTA, hence LA is never higher than 5 (the number of processes that are involved). > working a lot harder than I'd expect. For reference, the Windows > box I replaced was a DL380 G2, with a single P3-1.4 and 256mb RAM, > and it was running a commercial antivirus product based on Sendmail. Sendmail for NT? AFAIR that product is able to relay 1-2 (or was it 4?) messages/second. sendmail 8 itself (just the MTA on a Unix box with a disk subsystem that has a battery-backed cache) is able to relay about 500 msgs/s (depending on CPU, RAM, network I/O capacity, etc of course, and - most important - DeliveryMode). > What can I do to diagnose the performance bottleneck? The CPU is > mostly idle. So what drives the load up? Do you run several milters on the machine (each of which shouldn't contribute more than 1 to LA as those are multi-threaded)? PS: you might want to run some of those disk I/O benchmarks to determine the number of IOPs your system can provide.
