Red Hat Bugzilla – Bug 24234
Stopping squid sometimes result in an endless loop
Last modified: 2014-03-16 22:18:14 EDT
When squid died abnormally (or was killed with -9 or so) and left a pid file in /var/run, running "/etc/rc.d/init.d/squid stop" will result in an endless loop. The squid stop() routine checks $? after running "$SQUID -k shutdown &", but this will always give 0, because the process was started in the background. Given the wish to give some output while waiting for the shutdown, a counter can be included to let the loop only be passed a finite number of times. Another possibility is to first do a "squid -k check" synchronously, but this still leaves some room for an endless loop (at least as a race condition). Maybe both solutions should be combined. I'll attach a patch later.
Created attachment 7748 [details] Patch for squid init script.
Justification for the changes made in the patch: (1) The "-k check" is included to catch most cases, when squid is just not running anymore and the pid file still exists for some reason (e.g. kill -9). (2) The time counter is included, in case a squid process exists (assuming -k check works reliable), but won't shutdown for some reason. The timeout can be fairly long, as this will only happen in very rare cases (if ever), but it's still better than having to push the reset button :-).
Will be fixed in 2.3.STABLE4-7. The variable will be set from the new /etc/sysconfig/squid, however - thanks for the patch.