On Tue, Apr 02, 2002 at 05:11:19PM -0500, Duncan Findlay wrote:
| On Tue, Apr 02, 2002 at 01:00:12PM -0500, Craig Hughes wrote:
| > On Tue, 2002-04-02 at 04:27, Matt Sergeant wrote:
| > > One thing to beware of is that Perl will *never* free memory back to the
| > > OS. Ever.
| >
| > That shouldn't matter much in our case, because of the
| > spawn-die-spawn-die-spawn-die nature of spamd. The only process that
| > could be "growing" is the master spamd process, but it's never doing
| > anything that allocates memory.
|
| The only problems come when you get "spawn-spawn-spawn-spawn-spawn-spawn-
| spawn-spawn-spawn-spawn-spawn-spawn-spawn-spawn-spawn-spawn-spawn-spawn-
| spawn-spawn-spawn-spawn-spawn-spawn-spawn-spawn-spawn-crash-and-burn" as
| soon as fetchmail kicks in.
Have you ever seen a load average of 29.5? (I have, and fetchmail
wasn't involved though SA was)
That's why exim has the
deliver_queue_load_max
option. I have mine set so that if the load average is above 5.0,
then exim will not spawn a delivery process, which means that no SA
processes will be spawned. The message will just sit in the queue
until the next queue run when it will deliver it if the load average
is below 5.0.
-D
--
If any of you lacks wisdom, he should ask God, who gives generously to
all without finding fault, and it will be given to him. But when he
asks he must believe and not doubt, because he who doubts is like a wave
of the sea, blown and tossed by the wind.
James 1:5-6
_______________________________________________
Spamassassin-talk mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
