On Aug 27, 2013, at 6:11 AM, Golden Shadow <firas73...@yahoo.com> wrote:

> Hello Squid Users!
> 
> I have a TPROXY server built with squid V. 3.3.7, which was compiled from 
> source. The server has 192 GB of RAM and 2 X 2.7 GHz CPUs, each with 12 
> cores. but I'm running only one squid worker. At the moment, there are 17 
> squid clients and the server is receiving 20025 http requests per minute. CPU 
> usage is 62%. Squid is consuming about 12 GB of memory at the moment.
> 
> Traffic is being redirected to this TPROXY using WCCP. Each once in a while, 
> traffic redirection stops for about 5-10 minutes. When checking the WCCP 
> clients on my Cisco switch, I see there is no WCCP client! 
> 
> During this pause time, running "squidclient mgr:info" freezes endlessly and 
> does not return any results. Checking /var/log/squid/cache.log when this 
> pause happens does not show any strange logs.
> 
> When this condition happened last time, I shutdown squid using "service squid 
> shutdown" and the init script reported that squid shutdown went well but 
> running "ps aux | grep squid" returned the following:
> 
> root      6900  0.0  0.0 103244   852 pts/0    S+   12:05   0:00 grep squid
> root     13006  0.0  0.0  38964   392 ?        Ss   Aug23   0:00 squid -f 
> /etc/squid/squid.conf
> squid    13009 63.1 25.0 50332228 49622116 ?   Dl   Aug23 3383:07 (squid-1) 
> -f /etc/squid/squid.conf
> 
> As you see, the worker squid kid is in uninterruptible sleep state and it 
> remained in this state for a while. Even when squid seems to be working 
> properly, the worker process may be in uninterruptible sleep for a short 
> time, is this normal? 
> 
> Becuase squid worker process stayed in the uninterruptible sleep mode for 
> quite long, I tried to stop it using "squid -k shutdown", which failed with 
> the following:
> squid: No running copy
> 
> I'm wondering why this command is claiming "No running copy" while both the 
> parent and kid squid processes are running! 

Squid might be dumping core during that interval. Have you checked the 
cache.log and/or looked for coredump files in the coredump_dir?

Guy

Reply via email to