Hi, On Thu, Feb 14, 2008 at 09:16:17AM -0700, Damon Estep wrote: > > > > -----Original Message----- > > From: [EMAIL PROTECTED] [mailto:linux-ha- > > [EMAIL PROTECTED] On Behalf Of Dejan Muhamedagic > > Sent: Thursday, February 14, 2008 8:29 AM > > To: General Linux-HA mailing list > > Subject: Re: [Linux-HA] STONITH frustration... external/rackpdu > > > > Hi, > > > > On Wed, Feb 13, 2008 at 10:39:07PM -0700, Damon Estep wrote: > > > More details; > > > > > > Note:resource_cn2_stonith is the resource the shoots node cn2, > > resource > > > that should shoot cn1 is running on cn2 which is online. > > > > > > [EMAIL PROTECTED] ~]# crm_verify -LV > > > crm_verify[5162]: 2008/02/13_22:33:51 WARN: > > > determine_online_status_fencing: Node cn1 > > > (0048dc5f-f558-4ae4-a9bb-d0f62b0b4b5a) is un-expectedly down > > > crm_verify[5162]: 2008/02/13_22:33:51 WARN: determine_online_status: > > > Node cn1 is unclean > > > crm_verify[5162]: 2008/02/13_22:33:51 WARN: native_color: Resource > > > resource_cn2_stonith cannot run anywhere > > > > It says here that the stonith resource doesn't run. Did you try > > using a clone stonith resource? That should be better > > configuration for this type of stonith device. > > > > Thanks, > > > > Dejan > [Damon Estep] > > That is the stonith resource for the good node, running on the bad node. > The stonith resource for the bad node is called resource_cn1_stonith and > is running on cn2 at the time of that log entry. > > I have tried a clone, the stonith process is not able to determine which > node is capable of shooting the missing node. The current theory is that > the rackpdu plugin does not return a hostname it can shoot when queried, > and a review of the script confirms that it has no mechanism to return a > hostlist when queried.
Right, and that's a prerequisite for all stonith plugins. They must return (print) a list of nodes which they can manage. I'm afraid that this stonith plugin is not good. If anybody's using it, I'd really like to know how. The parameters are host/community/outlet, but they should be host (as in 'which node is managed here") ipaddr (the ip address of the stonith device unit) community outlet (the outlet to which this host is connected) Thanks, Dejan P.S. CC-ing the author, perhaps he could help... > If you see evidence in the logs to the contrary let me know. > After giving it some thought overnight I think I might take a shot a > building a generic snmpV1 plugin, so as long as you know the OID and > values for on|off|reset you should be able to use it with any snmp v1 > addressable power switch or management card. > _______________________________________________ > Linux-HA mailing list > [email protected] > http://lists.linux-ha.org/mailman/listinfo/linux-ha > See also: http://linux-ha.org/ReportingProblems _______________________________________________ Linux-HA mailing list [email protected] http://lists.linux-ha.org/mailman/listinfo/linux-ha See also: http://linux-ha.org/ReportingProblems
