[squid-users] issue with start / stop scripts

Amos Jeffries squid3 at treenet.co.nz
Sun Nov 29 02:56:43 UTC 2015


On 29/11/2015 11:32 a.m., Eliezer Croitoru wrote:
> A check on what?
> Basically to verify if squid is still running you need to verify that
> there are is not one squid instance running.
> The PID is kind of a hack to make sure squid is still there or not.
> In most cases you can cancel the timeout and check only for the PID.
> Also notice that there is a "rm -rf" there which was inherited from an
> old script that I got as a "gift" since my own script got lost in a
> server migration.
> 
> You can run three checks in parallel:
> - the pid exists or not
> - the process exists or not(using "ps aux|grep squid")
> - check if the port in netstat is still in listening mode.
> 


Also, be aware the shutdown signal for squid is SIGHUP. If you need to
abort the internal timeout use a second SIGHUP.

So the shutdown process as seen/done by external scripts should be:

* scan squid.conf for unusual pidfile_path and shutdown_lifetime values
* if squid.pid exists; send SIGHUP to process indicated inside
* else locate running squid process and send SIGHUP
* wait desired timeout + a few seconds

* if process still exists; repeat SIGHUP
* wait a few seconds more

* if process still exists; send SIGKILL; repeat until closed
* if squid.pid still exists; rm squid.pid


Notice the two points where waiting "a few seconds more" is needed for
the signals to have effects. That is the time Squid *actually* takes to
shutdown.

The shutdown_lifetime is a grace period where normal proxying operations
are still taking place to try and finish client transactions off. It is
essentially the *minimum* time needed for Squid to shutdown. If you just
abruptly abort/SIGKILL right at the same point where squid should only
be starting to seriousy wind up its own internal actions it can lead
cache and FD problems.

HTH
Amos



More information about the squid-users mailing list