>>> Lars Ellenberg <lars.ellenb...@linbit.com> schrieb am 26.03.2011 um 00:10 in Nachricht <20110325231023.GK24099@barkeeper1-xen.linbit>:
[...] > Pacemaker is not a substitute for proper monitoring (nagios, whatever). [...] AFAIK, pacemaker starts a monitor programm every now and then to check the state of a resource. In contrast HP ServiceGuard uses a monitoring process that runs forever. Depending on the complexity of the thing to monitor, that latter approach may have performance benefits. For example if you are going to monitor a bunch of processes to exist, you'll have to get their PIDs only once and then just check is they are there (using the system call, not the "ps" command). If you are starting the monitor every time, you are wasting resources. The less impact the monitoring task has, the more frequently you can monitor your resources. I'm no fan of "monitoring a system to death" (like causing a system load of 10% for monitoring). Back to the problem: If you write a program that continually monitors a resource, it's easier to detect (and possibly log) the state changes. Whether to actually care about such changes is a matter of taste. For pacemaker I can imagine some kind of "proxy" to be used as a monitor that is actually querying a monitor job that is continuously running. Regards, Ulrich _______________________________________________ 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
