Hi, On Mon 08 Aug 2011 06:38:13 PM CEST, Ritter, Nicholas wrote:

I am testing TPROXY with Squid v3.1.14 on CentOS v6 and noticed several TCP-performance related problems were squid is failing due to TCP session handling issues. CentOS v6 is running a 2.6.32-71.29.1 kernel, and I saw a TPROXY listserv posting from Nov 10, 2010 about TPROXY/Kernel hanging issues and a fix that was merged in and after 2.6.37.

My question is if this issue could be the root cause of the problems I am seeing, and if anyone had suggestions for further debugging and solutions. Squid is reporting a higher than normal number of 502 BAD GATEWAY errors and some websites consistently not working. I have testing the squid install without TPROXY and the errors go away, so I know it is TPROXY related.

If upgrading the kernel is the way to go, that gets hairy with CentOS and I am not sure which method is the best way to go for upgrading the kernel.

Any thoughts or suggestions?

Well, I'm not quite sure the issues you're experiencing are fixed by those commits in 2.6.37, but you could certainly give it a try. Based on the RedHat Bugzilla I think all those patches have been merged into the RHEL 6.1 kernel: https://bugzilla.redhat.com/show_bug.cgi?id=591335 I don't know if CentOS packages are available yet -- Scientific Linux already has a version of the kernel which contains the above patches. -- KOVACS Krisztian