Hello, Alexander Fisher wrote: > A word of warning. If the two PDUs can't communicate (perhaps a > switch is down or a bad cable) > and you fire the reset command at the PDU you can talk to, it'll > appear to have worked > (from the clusters perspective), but obviously won't.
I think this warning should be spelled in capital letters. Because this of course makes the whole feature unusable. I'm really surprised how fast this cool feature enters the recommendation of this cluster project without any investigation and testing. > I don't know how worried about this I should be. Should I be worried > about stonith working if I've > got a switch failure? you should only be worried if you care about your data. > I think I'll use IPMI and rackpdu in the same configuration. > If I connect both PDUs to the same switch and all IPMI ports to the > other, I *think* everything will be fine, > in the event of either a switch failure or a power failure leading to > one dead PDU and one dead switch. > > I think I need to draw a diagram! :) Not for me. You can connect the PDUs to as many switches as you like, the problem remains: stonith will return success even if the second PDU is not working properly, not responding, not connected at all, or for whatever other reason. Cheers, Peter _______________________________________________ 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