I added shiny new hosts to my ONe cluster and I am trying to phase out the old ones. The problem is that “onehost flush oldhost1” does not migrate any VM to the new hosts: in my sched.log, there is the following message for each VM running on oldhost1:
Match-making results for VM 1913:
Cannot schedule VM, there is no suitable host.
However, when I explicitly run “onevm migrate --live 1913 newhost1”, the VM gets migrated and is running corrrectly. Also when I create a new VM and make it explicitly runnig on a new host, it starts up correctly.
When I try to reschedule the VMs using “onehost flush oldhost1”, there is only the above error message added to sched.log, but nothing special to the VM log (/var/log/one/1913.log).
The new and old hosts are in the same cluster (id=101). I run ONe 5.6.1 on CentOS 7, with kvm/qemu-ev.
Where should I look for reasons why does the match-making fail? Thanks!
I guess with 1. and 3. I would not be able to migrate the VMs manually. The host capacity is easily visible in the Infrastructure->Hosts tab, so 2. is also not a problem. What helped was 4., I was able to see which rules the scheduler evaluates.
The problem was that the newly added hosts were not in the default VDC for the group “users”, in which most of the VMs are created. So, the problem was between the keyboard and chair after all. Sorry for the noise.
That said, it would be nice if scheduler could log the detailed output even on the default log level for scheduling actions which fail. Also, it would be helpful if Sunstone could display the suitable hosts for a given VM, for example when trying to manually migrate the VM: use a different color for hosts to which migration is allowed according to ACLs and capacity, and a different color where manual override is possible.