Have you tried the "operation system dependent wierdnesses" section in the FAQ?
Did the: echo 0 > /proc/sys/net/ipv4/tcp_ecn Did not help. Rebooted the Linux CentOS 4.4 box running Squid and the problem went away. Any ideas? This is a transparent cache with a router DST-NAT'ing all port 80 traffic to it. Has been working fine for months. Have a 2nd identical setup that has not exibited this problem. One odd thing about this server. Its an AMD64 dual core although not running 64 bit OS. The clock drifts severely. I do a ntp update every hour to compensate but it seems to drift about 20 minutes by that time. Could that cause it? The other setup with different motherboard but still AMD64 does not drift. I see this in logs once in a while. STALE: Entry's timestamp greater than check time. Clock going backwards? Not sure if this could cause anything? Thanks. Matt
