Re: [squid-users] Re: SQUID 3.1.10 keep 100% cpu after exiting

From: Amos Jeffries <squid3_at_treenet.co.nz>
Date: Fri, 28 Jan 2011 05:56:28 +1300

On 27/01/11 23:13, Sunny wrote:
> Hi there, can somebody suggest how should I provide more info?
>
> at that time squid does not response to squidclient anymore.
>
> Cheers.
>
> On Mon, Jan 24, 2011 at 8:28 PM, Sunny<sunyucong_at_gmail.com> wrote:
>> Hi there,
>>
>> since I upgraded to 3.1 (both .6 .9 and .10) have such problem.
>>
>> After I issue "squid3 -k shutdown", the squid will exit but the
>> process won't stop after it clams "exit normally"
>>
>> The process will still use 100% cpu and network connections are not
>> closed due to output of "netstat"

?? all of those "Open FD READ/WRITE" lines you see in the log are Squid
forcibly closing the FD sockets which were still open after the 30
second grace period. The OS itself may keep them in TIME_WAIT or similar
for a while after they are released.

>>
>> root_at_cache-1:/tmp/squid-3.1.10/errors/zh-cn# ps aux | grep squid
>> root 12557 0.0 0.0 32272 1180 ? Ss 19:56 0:00
>> /usr/sbin/squid3 -YCF -f /etc/squid3/squid.conf
>> proxy 12559 53.5 21.9 924084 889552 ? Rl 19:56 14:04
>> (squid) -YCF -f /etc/squid3 squid.conf
>> root 12691 11.0 0.0 8952 876 pts/0 S+ 20:22 0:00 grep
>> --color=auto squid
>>
>>
>> I'm running squid on a virtual machine ubuntu 10.04. Here attached the
>> exiting log.
>>
>>
>> 2011/01/24 20:18:08| Preparing for shutdown after 8692 requests
>> 2011/01/24 20:18:08| Waiting 30 seconds for active connections to finish
>> 2011/01/24 20:18:08| FD 11 Closing HTTP connection
<snip>
>> 2011/01/24 20:20:17| Open FD READ/WRITE 353 minigameimg.qq.com idle connection
>> 2011/01/24 20:20:17| Open FD READ/WRITE 354 Waiting for next request
>> 2011/01/24 20:20:17| Open FD READ/WRITE 357 Waiting for next request
>> 2011/01/24 20:20:18| Squid Cache (Version 3.1.10): Exiting normally.
>>

Last message Squid prints before it releases all its remaining memory
and stops. Your ps trace indicates there are about 1 or 2 GB of RAM in
use by the globals that needs dumping.

Could you run a few such ps over a minute or so and see if the RSZ and
VSZ are decreasing? It could just be taking a long time for the OS to
free all that RAM.

The "strace" debug tool can hook into a running binary and display what
its doing. The second binary on your list named "(squid)" is the one to
watch with this.

Amos

-- 
Please be using
   Current Stable Squid 2.7.STABLE9 or 3.1.10
   Beta testers wanted for 3.2.0.4
Received on Thu Jan 27 2011 - 16:56:33 MST

This archive was generated by hypermail 2.2.0 : Thu Jan 27 2011 - 12:00:03 MST