Joe, The way I'd fix this problem would be to change the stonith action to poweroff, then I would bring each machine up in single user mode and make sure that you won't automatically join the cluster when the machine comes up (chkconfig openais off in my system, ymmv.) Then you can bring it up to runlevel 3, clear out any directories or use whatever tools you need to manually blow away the configuration and then start openais to rejoin the cluster. ------
That being said, there's probably some magic you can do with the crm node commands that will do the above better. ------ In the future, using the crm shell, try this: app-03:~ # crm crm(live)# configure crm(live)configure# edit ... which should put you in an editor mode (whatever your $EDITOR variable is set to, mine's vim) which should make it easy to remove any resources and do major configuration surgery. Don't forget to validate and commit after editing. ------ Keep in mind that various crm and lrm bugs might make things a touch wobbly with big configurations. I've found that in some cases when I screw up a configuration big time, I have remaining LRM status entries in the XML file that actually contains the configuration. Running `crm resource cleanup` has helped, er, clean these up. Disclaimer: I'm still in the testing phase, I'm using the shell 100% and not the GUI, I'm using SLES11-HAE, and we're only running some noncritical systems in testing mode. I will accept no blame if the above causes the heat-death of the universe or a black hole to form in the center of the planet. -K --- Karl Katzke Systems Analyst II TAMU - DRGS _______________________________________________ Pacemaker mailing list Pacemaker@oss.clusterlabs.org http://oss.clusterlabs.org/mailman/listinfo/pacemaker
