On 27/10/2013 2:54 a.m., Omid Kosari wrote:
After upgrade to 3.3.8 from 3.1.20 the cpu usage of squid grows triple times without change to config . Please look at attached image of cacti graph <http://squid-web-proxy-cache.1019090.n4.nabble.com/file/n4662906/cpu_usage.png> my configs are available in following posts http://squid-web-proxy-cache.1019090.n4.nabble.com/Squid-naps-each-3600-seconds-tp4662811.html http://squid-web-proxy-cache.1019090.n4.nabble.com/rock-questions-tp4662816.html http://squid-web-proxy-cache.1019090.n4.nabble.com/rock-problem-on-each-squid-restart-tp4662864.html
Is workers enabled on the 3.3 version? or exactly the same 3.1-compatible non-SMP confg file?
* 3.1 version cannot consume more than 100% on a single core. Which in multi-core systems shows as a low % of total CPU. 3.3 with SMP workers can use up to 100% on N cores, which shows up as up to 100% of total CPU.
* rock store scans the storage DB file to load cache index. This can take a lot more CPU immediately after startup than other disk storage types do. 3.1 does not supprot rock and so will not show that behaviour.
Amos
