Cute. I've used that feature in sendmail.
/proc/loadavg also contains the load average and should be faster to read than running uptime.


Peter Eisch wrote:
I've goofed around trying to find the right number for managing the tcp
connections on the various systems for too long.  Today I crumbled and took
a lame way out.  So....

Announcing 'loadcheck' which will look at the current 1 minute system load
and perhaps decline taking a transaction if we're already busing doing
things.

If I let too many connections dump off to spamassassin at the same time it
might go nuts and take minutes to return.  This is an effort to keep the
system responsive while still allowing as many connections as it can
expediently handle.

peter

Reply via email to