On Wed, Nov 30, 2011 at 9:46 AM, <alain.mou...@bull.net> wrote: > Hi > > I 'm facing a problem with this RA, I wonder if is it a real one already > identified : > in fact, the stop of the mysql resource fails and therefore the node is > fenced. > I've checked a little in the RA script : the stop kills the pid retreived > from > /var/lib/mysql/mysqld.pid > as indicated in my params : > params binary="/usr/bin/mysqld_safe" pid="/var/run/mysqld/mysqld.pid" \ > but this daemon has /usr/bin/mysqld_safe as father, as this mysqld_safe > is expected to restart automatically the mysqld daemon if it is killed of > if > it fails. > So the stop target of the RA kills the daemon mysqld but as it is > instantaneously > restarted, it proceeds to SIGTERM, then SIGKILL, etc. until timeout on > stop > and it is logical because the daemon is restarted just after each kill > > So is there something I miss in my configuration ? > Or is this RA not well working when using /usr/bin/mysqld_safe as binary ?
This default is a leftover from the old heartbeat v1 days, when the cluster manager didn't do any resource monitoring on its own, and starting MySQL under its angel process was more than helpful. This is now obsolete. We've had a few discussions about changing this default here on the list, but those always died in the end because sadly, we have no real good way of alerting users to changed defaults in RAs. In a Pacemaker configuration with resource monitoring enabled, you ought to just use mysqld directly. If you're on RHEL or a derivative, this means that you'll have to set the "binary" parameter to the full path for mysqld, which on that platform (somewhat strangely) lives in /usr/libexec which is not in the default $PATH. If you're on a different platform you should be able to just use "mysqld" as that would be found in /usr/sbin there. Cheers, Florian -- Need help with High Availability? http://www.hastexo.com/now _______________________________________________ Linux-HA mailing list Linux-HA@lists.linux-ha.org http://lists.linux-ha.org/mailman/listinfo/linux-ha See also: http://linux-ha.org/ReportingProblems