[squid-users] pushed to the limit

greetings,

I have a faithful squid box that has been hanging under extreme loads. I don't suspect squid as a failure but some theoretical limit.
I have 2.8 Ghz Quad-core with 4G 800mhz ram, running squid 3.1.11 on osx

squid is configured to intercept port 80 only and allow passthrough no cache with a redirect to squid guard.
 fwd 10.0.1.1,3128 tcp from 10.1.0.0/16 to any dst-port 80 in recv en1

gigabit inbound en0
gigabit outbound en1

during the peak the box is doing in excess of 132Mb/s with 17,000 tcp connections, the same time each day, the box just hangs.
squid is still running but nothing going in or out. logs don't show anything out of the ordinary.the router log shows the interface going down and then coming up. the only thing I notice is high traffic.
kick start the box, restart squid and everything is back to happy.
because we are seeing it at the same time, I do suspect a rouge process that should be performed at another time. However,…

because of the age of the system I'm running this squid on ( osx 10.5.8 squid 3.1.11 ) , I also suspect some kernel limitations that I am exceeding.

I have made one adjustment, I have set my ipfw fwd statement from my largest areas with a tcp limit,

$cmd 508 fwd 10.0.1.1,3128 tcp from 10.15.0.0/16 to any dst-port 80 in recv en1 limit src-addr 10
and as well for anything else,
$cmd 1615 allow ip from 10.15.0.0/16 to any limit src-addr 10

if anyone has any performance tweaking suggestions.
I do have plans to add another box and split the load, but in the mean time, I need to locate what is causing this hiccup and alleviate it if i can.

thanks for any input
-j
Received on Fri Dec 13 2013 - 04:25:49 MST