Hi Amos , thanks for reply Regarding to your description If I have 10K ips with 10k listening ports ...... Will each squid process handle 10 K ? Or I need to distribute the ips/ports to each process ??? cheers -----Original Message----- From: Amos Jeffries [mailto:squid3@xxxxxxxxxxxxx] Sent: Friday, September 25, 2015 12:44 AM To: Ahmad Alzaeem Subject: Re: squid with SMP registeration time out when i use 10K opened sessions On 25/09/2015 9:27 a.m., Ahmad Alzaeem wrote: > Hi amos > I have alredy 5.2 3.5.9 is the latest security release. > > All squid traffic go to one of cpus and it load it 100 % I don’t care > caching or other thing > > All I need is to load like 20K ports and be balanced to the cores You wont get balance on the cores. They actually work *better* when unbalanced. It makes the core L2/L3 RAM cache work with closer to optimal contents. With 3.5.7 or later run several Squid with: squid -n squid1 squid -n squid2 squid -n squid3 ... The "${service_name}" variable in suqid.conf will become that "squid1", squid2" "squid3", etc You need to make your squid.conf contain: include /etc/squid/${service_name}-ports.conf pid_filename /var/run/squid/${service_name}.pid cache_log /var/log/squid/${service_name}-cache.log ... and so on for the other directives listed in <http://wiki.squid-cache.org/MultipleInstances> Make the squid1-ports.conf etc config files listing your http_port's and a unique_hostname for each Squid instance. Maybe access_log as well. Also set cpu_affinity for each instance to tie them to different cores. Amos _______________________________________________ squid-users mailing list squid-users@xxxxxxxxxxxxxxxxxxxxx http://lists.squid-cache.org/listinfo/squid-users
