[squid-users] squid with SMP registeration time out when i use 10K opened sessions

Ahmad Alzaeem ahmed.zaeem at netstream.ps
Tue Sep 29 18:32:51 UTC 2015


Hi amos I did , but can u tell me why that happened ??
Look @ my cpu cores for squid when it has high traffic :

  1  [####*********************************************100.0%]     Tasks: 29, 4 thr; 6 running 
  2  [                                                   0.0%]     Load average: 2.53 0.86 0.36 
  3  [                                                   0.0%]     Uptime: 2 days, 07:21:15
  4  [*                                                  0.7%]
  Mem[||||||||#*                                   569/3950MB]
  Swp[                                                0/255MB]

  PID USER      PRI  NI  VIRT   RES   SHR S CPU% MEM%   TIME+  Command                                                            
12350 squid      20   0  261M 77512 12488 R 20.0  1.9  4:04.68 (squid-1) -n squid1                                                
12367 squid      20   0  265M 80736 12548 R 20.0  2.0  4:31.87 (squid-1) -n squid3
12360 squid      20   0  259M 74840 12468 R 20.0  1.9  4:04.30 (squid-1) -n squid2      
12384 squid      20   0  276M 92584 12532 R 20.0  2.3  5:35.10 (squid-1) -n squid4
12478 squid      20   0  259M 74724 12468 R 20.0  1.8  3:56.48 (squid-1) -n squid5
16650 root       20   0  110M  3324  2524 R  0.0  0.1  0:02.83 htop               



As you see , the 5 instances has about same loading , but it seems  the cpu mapping is not working ok !!!

I did add the command 
process_numbers=1,2,3 ,4,5cores=1,2,3,4,5

to each instance of squid !!

what coud me doing wrongh ?

or how to test if there process got the cpu mapping as ok or not

thanks a lot 

-----Original Message-----
From: Amos Jeffries [mailto:squid3 at treenet.co.nz] 
Sent: Monday, September 28, 2015 11:59 PM
To: Ahmad Alzaeem
Cc: squid-users at lists.squid-cache.org
Subject: Re: [squid-users] squid with SMP registeration time out when i use 10K opened sessions

On 28/09/2015 7:24 p.m., Ahmad Alzaeem wrote:
> Hi amos
> 
> I have 10 K
> 
> I DIVIDED them to 5 files
> 
> Each file has 2 K
> And each file has its own cache.log file /visible name ....etc
> 
> The question im asking is :
> 
> Do I need to put the  directive in  cpu_affinity_map 
> process_numbers=1,2,3 ,4,5cores=1,2,3,4,5 In squid.conf ??
> 
> Or I need to go to each separated file  of each instance and provide 
> command there

Each separated instance file needs a different mapping to use your 15+ CPU cores.

Think about it: you are dividing cores up between Squid instances.

Amos




More information about the squid-users mailing list