On Fri, 5 Mar 2010, RW wrote:
On Fri, 5 Mar 2010 10:21:28 -0500 (EST)
James Smallacombe <u...@3.am> wrote:
I actually meant spamc's that had no spamd children to talk to. It
seems the ports version of SA replaced my original sa-spamd startup
script, which raised the max children from (default?) of 5 to 30.
sa-spamd doesn't define the number of children, if you want spamd to
start with non-default settings you can define spamd_flags in rc.conf
I had trouble with that line in the rc.conf with some older version of SA
long ago, so put it in local/rc.d/sa-spamd. As you noted, this burned me
because installing a new port of SA overwrites that file, which didn't
occur to me ahead of time. Oops.
just reinstalled 3.3.0 from source instead of ports
Sounds like a bodge to me, and presumably it's using perl 5.8 because
you replaced perl not because you installed SA manually.
See my previous email...even after deinstalling perl 5.10 and cleaning SA
in ports, it would still list 5.10 as a dependency...I'm not entirely sure
how the SA port determines what version of perl is already installed, if
any.
and fixed the startup script,
By the sound of it, your hacking the script was the only actual problem,
so why repeat it.
Good point...I will probably go back and re-try putting the flags where
they belong in /etc/rc.conf and see how it goes.
so now it uses perl 5.8.9, yet it still seems to be
using a lot more CPU than 3.2.5 did...is this normal?
load averages: 5.76, 5.19, 4.19
load averages: 3.20, 4.60, 4.15
By the look of it you are using way too many spamd children.
Actually, I didn't have enough with 5. I posted this a couple of minutes
after raising it to 30, but I should have just waited several more
minutes; things calmed way down and look fine now. It was using well over
a dozen children for a while, and now is at just a few.
Note that I did not raise the min children, only the max. Trouble really
starts when it hits the max.
James Smallacombe PlantageNet, Inc. CEO and Janitor
u...@3.am http://3.am
=========================================================================
