On 21/05/2014 4:10 p.m., Cinaed Simson wrote:
> On 05/20/2014 08:46 AM, fernando wrote:
>> Hi,
>>
>> I don't know if this affects only the RPM packaging or squid sources. I'm
>> using Eliezer packages for CentOS 6, namely 3.4.3-1.el6.x86)64.
>>
>> The issue is that sometimes "service squid restart" leaves old squid
>> processes running alongside newer ones. I have to "killall squid" and start
>> again to have a working proxy.
>>
>> I see "squid -k shutdown" does not wait for all squid kids to stop, and
>> this makes init.d/squid to leave such broken restart.
>>
>> I coud just increate SQUID_SHUTDOWN_TIMEOUT at /etc/sysconfig/squid so the
>> init.d/squid script wait a little longer, but i think the restart command
>> should be more robust. So I propose the following simple change to the
>> init.d/squid script:
>>
>> ---------------
>> restart() {
>> stop
>> RETVAL=$?
>> if [ $RETVAL -eq 0 ] ; then
>> rm -rf $SQUID_PIDFILE_DIR/*
>> start
>> else
>> echo "Failure stopping squid or stopping squid took too long. Please check
>> before restarting."
>> return 1
>> fi
>> }
>> ---------------
>>
>> Instead of blindly calling start after stop, check if the stop function
>> returned ok or an error status code.
>>
>> I think it's better having restart fail sometimes than having broken squid
>> processes after restart.
>>
>> []s, Fernando Lozano
>>
>
> The problem as I see it is there needs to be timeout (or a wait) between
> stop and start in restart.
>
> The stop branch returns immediately and then executes the start branch
> which sees the pid file (since the controlling squid is waiting for the
> children to complete their processes) and claims squid already running
> and exits.
>
> So when the controlling squid finally stops and removes the pid file the
> start branch has already exited and the server doesn't restart.
>
> I'm not an expert on squid but I would not recommend removing the pid
> file before squid controller has exited.
>
> If it never stops then you have different problem.
The problem seems to be that the script is trying to jig up its own
sequence when squid -k restart provides a well tested and standard
method already.
Amos
