Thanks- that is good to hear that you've seen heavier-load installations than mine that don't exhibit the problem. Were any of them running on Linux x86_64? Maybe that's the culprit. There must be some combination of factors.
It's just strange because I don't think there's anything all that weird about my config, and I have run a number of different configs (i.e. with & without a disk cache, with & without an IP-based access ACL, different squid versions, CentOS package and built from source, etc.) The leak didn't start becoming a problem until the load on the servers got really high, so as I said I think there really is a leak in there somewhere and if more servers were handling high traffic loads there would probably be more talk about it. Maybe squid on other OSes like BSD does not have this problem, maybe it's specific to 64-bit, I don't know. I've already increased the max file descriptors- my squid started crashing about a year ago when it began running into CentOS's default limit so I fixed it (squid did a great job of logging the imminent crash from file descriptor exhaustion btw). The cache and access logs aren't the problem- they rotate daily and never get to be all that big. Just wish I had some squid development experience so I could easily get into a debugging environment and track this down... but squid's a big package and I know there will be a big learning curve to start debugging builds. Thanks -Ty
