On 11.10.2021 10:15, Ulrich Windl wrote: >>>> Andrei Borzenkov <arvidj...@gmail.com> schrieb am 10.10.2021 um 16:52 in > Nachricht <b2dd249c-a887-8e53-3010-1c2852525...@gmail.com>: >> On 10.10.2021 14:29, martin doc wrote: > > ... >> For each resource pacemaker computes allocation scores for each node >> (taking into account colocation). Node with the highest score wins. If >> multiple nodes have the same score, pacemaker should select node >> according to balancing strategy. > > From the output of "crm_simulate -LUs" it's very hard to see that. My guess > was that utilization even influences the initial score.
I do not see it. > For example: > > Allocation scores and utilization information: > ... > native_assign_node: prm_stonith_sbd utilization on h16: > pcmk__clone_allocate: cln_DLM allocation score on h16: 18060 > pcmk__clone_allocate: cln_DLM allocation score on h18: 15030 > pcmk__clone_allocate: cln_DLM allocation score on h19: 15030 > ... > > My guess it that the score from above is from utilization. > May be. Without seeing your actual configuration there is no way to even attempt to guess. What I see is normal scores according to (co-)location constraints augmented by utilization (i.e. resource would not come up anywhere if the only node it is allowed to run on would become overloaded). I have just two node test cluster, and resources are distributed across two nodes whatever strategy I set (with obvious exception of "minimal"). But as I already said in another mail, I do not quite understand the configuration and it was never fully described. _______________________________________________ Manage your subscription: https://lists.clusterlabs.org/mailman/listinfo/users ClusterLabs home: https://www.clusterlabs.org/