Hi Mr. Henrik,
2010/2/23 Henrik Nordström <henrik@xxxxxxxxxxxxxxxxxxx>:
>> We then used iptables to direct each network to a different
>> http_port (all with tproxy), and the time improved on the http_ports
>> that have fewer users...
>
> Check your /var/log/messages... most people running into problems like
> yours is running out of iptables conntrack entries
/var/log/messages is clean. No conntrack problem.
We had these in the very early stages of the squid deployment, but
changing the max solved it:
echo "524288" > /proc/sys/net/ipv4/netfilter/ip_conntrack_max
I'll try 3.HEAD code like mr. Jeffries suggested to see if it helps.
Any other thoughts beside that?
Thank you in advance,
Felipe Damasio
