Matthew Harrell <[EMAIL PROTECTED]> wrote:
>Can I get some pointers as to why I'm only getting somewhere between 25 and 50
>concurrent deliveries even though I have 120 in concurrencylocal. Attached is
>the qmailanalog output. Basically, I seem to get horrible transfer rates on
>this machine and I'm trying to figure out why.
>From our discussions earlier this month, it looked like your systems
were simply doing what they could. qmail can only run 120 deliveries
if there's sufficient CPU and I/O and network bandwidth. And, of
course, it's got to have sufficient pending deliveries to non-tcpto'd
hosts.
Things to look for:
high "so", "cs", in vmstat
low "id" in vmstat
high load average
high "bi"/"bo" in vmstat
insufficient network bandwidth
lots of deliveries to tcpto'd hosts
poor nameserver latency
What constitutes "high", "low", "insufficient", "lots", etc. isn't
carved in stone. I suggest you find a group of experts for your OS
(Linux, if I remember correctly) and work with them. This really isn't
a qmail-specific problem.
-Dave
