On 8/13/07, matilda matilda <[EMAIL PROTECTED]> wrote: > >>> "Andrew Beekhof" <[EMAIL PROTECTED]> 13.08.2007 15:14 >>> > > We will ensure that: > > a) all STONITH resources that can be active, are active before we > > issue a STONITH request > > I really don't understand this, sorry. I thought, that on > every node of the cluster the stonithd is started which > starts the stonith RA and is responsible for delegating > the stonith action to a node/ressource which is able to bring > the specified node down. Or do I have a total misunderstanding? > > So, in my case: When one node gets crazy the other node must be in charge > to stonith the crazy node, isn't it. But how would you achieve that > in a multinode cluster. Am I responsible to spread the right stonith RA? > > > b) all regular resources that need to be stopped, are stopped before > > we start stopping STONITH resources. this protects you if a resource > > fails to stop when you're bringing down the whole cluster. > > What does it mean to "stop a STONITH resource"? To shoot or not to shoot? > I'm sure I do need some explanation of terms.
a stonith resource is a way to configure the stonith daemon starting one tells stonithd to load a plugin with the supplied configuration stopping one tells stonithd to unload the plugin once the stonith resource is active, any machine in the cluster partition can make use of it. so what the changes mean is that we ensure stonith resources are active before we make any requests to stonithd (to shoot nodes) and that "normal" resources will be stopped (meaning we're unlikely to need to shoot any more nodes) before any stonith resources are does this help? _______________________________________________ Linux-HA mailing list [email protected] http://lists.linux-ha.org/mailman/listinfo/linux-ha See also: http://linux-ha.org/ReportingProblems
