On 11/29/2011 09:16 PM, Tim Serong wrote: > On 11/29/2011 04:28 PM, Muhammad Sharfuddin wrote: >> On Mon, 2011-11-28 at 21:47 +0100, Tim Serong wrote: >>> On 11/28/2011 06:54 PM, Muhammad Sharfuddin wrote: >>>> is it good/required to create order constraint for sbd resource >>>> >>>> I am using following fencing resource: >>>> >>>> primitive sbd_stonith stonith:external/sbd \ >>>> meta target-role="Started" \ >>>> op monitor interval="3000" timeout="120" \ >>>> op start interval="0" timeout="120" \ >>>> op stop interval="0" timeout="120" \ >>>> params >>>> sbd_device="/dev/disk/by-id/scsi-360080e50002377b8000002ff4e4bc873" >>>> >>>> I have following order constraints: >>>> >>>> order resA-before-resB inf: resA resB symmetrical=true >>>> order resB-before-resC inf: resB resC symmetrical=true >>>> >>>> should I also create another constraint for sbd like: >>>> >>>> order sbd_stonith-before-resA inf: sbd_stonith resA symmetrical=true >>>> >>>> please help/suggest. >>> >>> No. The STONITH resource doesn't need to be running in order for your >>> other resources to be operable (hence no need for an order constraint). >>> >>> Regards, >>> >>> Tim >> true, but if there is an order constraint for the STONITH resource, then >> it will at least make it sure that no other resource will be start >> before the STONITH resource. >> >> e.g: >> order sbd_stonith-before-resA inf: sbd_stonith resA symmetrical=true >> >> order resA-before-resB inf: resA resB symmetrical=true >> order resB-before-resC inf: resB resC symmetrical=true >> >> Because I stopped all the resources including STONITH resource(and >> stopping any resource sets the 'target-role="Stopped"'), then started >> all other resources else/except the STONITH resource, so at that time my >> cluster has no fencing resource available. > > So, don't stop the STONITH resource :)
And if you do it you are on your own ... you want it? you get it! :-) There are a lot of ways for an administrator to lower service downtime ... > > Side point: if you use "crm configure property stop-all-resources=true", > this will stop all resources *except* for any STONITH resources. The > point being, you do always want them running... > >> So in order to protect the cluster I thought that there should(must) be >> an order constraint that specifies that no other resource(s) will be >> start if STONITH resource is stopped/unavailable. >> >> Please suggest/recommend > > You should generally be OK without order constraints on STONITH > resources. I don't recall seeing any other systems where people had > created these constraints. I should also note that if, say, your > STONITH resource is running on node-0 and that node dies, the cluster > will start the STONITH resource on node-1, to kill node-0. It's smart > enough. > > Worst case, if your STONITH resource is completely broken, and a node > fails and thus can't be killed, the cluster will sit there and log > errors to syslog about its inability to kill the misbehaving node. > > (Question for everyone else: did I miss anything?) IIRC stonith resources are always started first and stopped last anyways ... without extra constraints ... implicitly. Please someone correct me if I'm wrong. Regards, Andreas -- Need help with Pacemaker? http://www.hastexo.com/now > > Regards, > > Tim
signature.asc
Description: OpenPGP digital signature
_______________________________________________ Linux-HA mailing list [email protected] http://lists.linux-ha.org/mailman/listinfo/linux-ha See also: http://linux-ha.org/ReportingProblems
