Hello list, I have a question regarding memory and CPU usage change from 3.0 to 3.1. I have 4 forwards proxies with ICAP (c-icap and clamav), NTLMv2 authentication, all four proxies each have about 200-400 req/sec on RedHat AS5 64bit servers with each 16GB mem for about 15k to 30k users. cache_mem is set to 3.5 GB, no disk cache, some ACLs are used. With 3.0.STABLE23 the typical memory usage was about 7 GB and CPU usage of about 30-40% for the main squid process, without the helpers or other processes. With 3.1.4 the proxies use up to 2GB more memory and CPU usage rose at times to 60% or 80%. After a restart of one of the proxies it behaved "normally" since then. Status: (the CPU usage, Client HTTP out and requests per second are all from last 5 minute-info) Proxy1: 36% CPU 7.2 GB mem 3.3 MB/s client out 341 req/s Version 3.1.4 Proxy2: 77% CPU 8.6 GB mem 2.6 MB/s client out 384 req/s Version 3.1.4 <----- Proxy3: 53% CPU 8.8 GB mem 6.8 MB/s client out 402 req/s Version 3.1.4 Proxy4: 32% CPU 6.8 GB mem 2.7 MB/s client out 405 req/s Version 3.0.STABLE23 Configuration of the proxies is the same, with only the changes for 3.1, mainly icap config. Is there some kind of memory leak that additionally causes massive CPU usage, could it be load related or is this normal behaviour? Martin
