On Wed, Oct 21, 2009 at 12:41 PM, <[email protected]> wrote: > Hello > > I am running an active - active cluster (mutual takeover) on RedHat 5. > - 2 Nodes, shared storage > - heartbeat-3.0.0-33.2 > - pacemaker-1.0.5-4.1 > - stonith > - 2 Oracle Databases for development and production, in failover state both > runnung on one node. > > I configured a group for a "base", including ip and mountpoints > and > a group including Oracle Listener and database. > I also configured, orders, preferred locations on this groups and all working > fine (mostly). > > Twotimes I had a ugly situation in failover -state,
What type of failure? > because the node running the basegroup before, doesn't stop Oracle, doesn't > unmount the Oracle filesystems and the second node mounts it anyway. \ Do you have ordering constraints between the various groups? This is how you tell the cluster which order to start/stop things in. > It is for Oracle databases a disastrous situation, because the filesystems > becomes partial distroyed and the databse is in a inconsistent state and > needs a recovery of the whole datbase. > Maybe it is a Split Brain situation and stonith should switch of the other > node, and I am sure the solution is a better configuration of the cluster. Do you have stonith configured? If not, this might be worth reading: http://theclusterguy.clusterlabs.org/post/113230399/highly-available-data-corruption > Anyway - I am searching with this question a "double bottom", means a second > check, if the filesystems still mounted on the other node. I wrote a script > for this, but where should I start it? Is there any "precheck" for a resource > where I can start a custom script? A precheck for the resource "filesystem"? > > Alternative solutions are welcome. So I set the meta attribute > multiple-active "stop_only" for all filesystem mounts, but it doesn't work in > that situation. > > Thanks > _______________________________________________________________ > Neu: WEB.DE DSL bis 50.000 kBit/s und 200,- Euro Startguthaben! > http://produkte.web.de/go/02/ > > > _______________________________________________ > Pacemaker mailing list > [email protected] > http://oss.clusterlabs.org/mailman/listinfo/pacemaker > _______________________________________________ Pacemaker mailing list [email protected] http://oss.clusterlabs.org/mailman/listinfo/pacemaker
