[squid-users] about cpu status

[Alex Rousskov]

On 07/14/2016 08:17 PM, johnzeng wrote:

> i deployed a set of squid 3.5.2 and redirect 300Mbps http traffic .
> 
>                 workers 4
>                 cpu_affinity_map process_numbers=1,2,3,4 cores=1,3,5,7
> 
> 
>                 i check cpu status via top command ,and i found one of
>                 squid is 92% at a time, whether it will be correct ?
> 
> 
> Tasks: 292 total, 5 running, 287 sleeping, 0 stopped, 0 zombie
> Cpu(s): 5.8%us, 1.2%sy, 0.0%ni, 89.4%id, 2.1%wa, 0.0%hi, 1.5%si, 0.0%st
> Mem: 65889452k total, 64218068k used, 1671384k free, 138480k buffers
> Swap: 29116412k total, 0k used, 29116412k free, 53204068k cached
> 
> PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
> 1982 cacheuser 20 0 33.9g 1.3g 408m R 92 2.0 228:49.60 squid
> 1983 cacheuser 20 0 33.9g 1.3g 408m R 56 2.1 254:23.07 squid
> 1981 cacheuser 20 0 33.9g 1.3g 406m S 36 2.0 159:40.23 squid
> 1984 cacheuser 20 0 33.5g 897m 408m R 33 1.4 189:14.00 squid


The uneven distribution of work among workers [on Linux] is discussed at
[1]. AFAIK, there were no significant relevant changes in Linux and in
Squid since that wiki page section was written, but YMMV.

[1]
http://wiki.squid-cache.org/Features/SmpScale#Will_similar_workers_receive_similar_amount_of_work.3F


HTH,

Alex.