On 10/04/2013, at 8:03 AM, Andrew Beekhof <and...@beekhof.net> wrote:
> > On 10/04/2013, at 4:33 AM, Felix Zachlod <fz.li...@sis-gmbh.info> wrote: > >>> It seems no rule has been selected here and it has fallen back to the >> default. >>> It looks similar on the other node- but some times it seems that the >> script is >>> being called with the correct params (without changing the cib, of >> course). I >>> have no idea, why the crm seems to randomly select one of the >>> configurations. >> >> Okay I found out what the problem was. The CRM first runs a monitor >> operation on the RA, BEVOR it has selected a node. This fails as it passes >> the default parameter and my validate_all function returns an error as it >> does not accept the dummy param. If the resource gets started for some >> reason on one node, the crm passes the param selected fort he 'running' node >> to BOTH RA on BOTH nodes- which fails on the node that had not been >> selected. >> >> I now added a param to my validate function to disable the HW check fort he >> monitor op. Now the resource works as expected (get the right params fort he >> node that has been selected by the crm). >> >> That is a bid odd behaviour from my POV. I think it would be a better >> solution to pass the parameters according to the node the resource is >> monitored on and not according to the node the resource is RUNNING on. >> >> So monitor resource on node B? Pass params adequate for node b! > > Absolutely! This is the intended behavior. Fixed in: https://github.com/beekhof/pacemaker/commit/c3e3922 Thanks for letting me know it was broken! _______________________________________________ Pacemaker mailing list: Pacemaker@oss.clusterlabs.org http://oss.clusterlabs.org/mailman/listinfo/pacemaker Project Home: http://www.clusterlabs.org Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf Bugs: http://bugs.clusterlabs.org
