Hi Alain, could you doublecheck, if the effect in your second test also happens, when you set a stickness/default-stickyness tobe something like 1000?
In your case when NOT setting default stickyness a running resource does not get a higher score than a stopped one, which is requested to be started. So hopefully (not tested, but my guess) in your second test case setting default stickiness will score up the running ressources and than also prevent resource3 to start, when conflicting resources resource1 and resource2 are already running and so all available hosts are "claimed". Fabian Von Samsung-Tablet gesendetMoullé Alain <[email protected]> hat geschrieben:Hi, a behavior which is not clear for me : 1/ Let's say we have 2 nodes node1 & node2 in the HA cluster, and 3 Dummy resources : resname1, resname2, resname3 and the forbidden colocation set like this : colocation forbidden-coloc-resname1-resname2 -inf: resname1 resname2 colocation forbidden-coloc-resname1-resname3 -inf: resname3 resname1 colocation forbidden-coloc-resname2-resname1 -inf: resname2 resname1 colocation forbidden-coloc-resname2-resname3 -inf: resname3 resname2 In this case, if resname1 is started on node1 and resname2 is started on node2, if we ask to start resname3, it does not start, and that 's seems correct for me because of both -inf: resname3 resname1 and -inf: resname3 resname2 Now, if the forbidden colocation are set like this : colocation forbidden-coloc-resname1-resname2 -inf: resname1 resname2 colocation forbidden-coloc-resname1-resname3 -inf: resname1 resname3 colocation forbidden-coloc-resname2-resname1 -inf: resname2 resname1 colocation forbidden-coloc-resname2-resname3 -inf: resname2 resname3 In this case, if resname1 is started on node1 and resname2 is started on node2, if we ask to start resname3, it does at first stop resname1, then migrate resname2 on node1, and finally start resname3 on node2 2/ Another try with two Dummy resources resname1, resname2 and the forbidden colocation set like this : colocation forbidden-coloc-resname1-resname2 -inf: resname1 resname2 If we ask to migrate resname2 to node1 , resname2 is stopped, resname1 is migrated to node2, and finally resname2 is started on node1. Now, the same test but with the forbidden colocation set like this : colocation forbidden-coloc-resname2-resname1 -inf: resname2 resname1 If we ask to migrate resname2 to node1 , nothing happens, resname1 remains on node1 and resname2 on node2 So, this seems to mean that the order of the resources for a -inf: collocation is important and has an impact on the behavior. I wonder if it is a normal behavior ? and so we have to really take in account the order on -inf colocation constraints ? or if there is a bug around this? Thanks Alain _______________________________________________ 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
