I upgrade my openstack controller node for nova-scheduler / api / conductor to newton and set the upgrade_levels to auto. After the upgrade i check the status für the nova-compute services on the compute nodes and they are shown as down.

After i check the logs for the controler nodes i found that the conductor seems to be able to communicate with the nova-compute service

compute-api1 nova-scheduler: 28031: INFO nova.scheduler.host_manager [req-45af962f-2816-4426-883f-0143b0bfc9a8 - - - - -] Successfully synced instances from host 'compute2'.

but also i see:

Seems service compute:compute2 is down is_up /usr/lib/python2.7/dist-packages/nova/servicegroup/drivers/mc.py:66

I test to upgrade one of the compute nodes to newton and it works. So it seems that the controller rpc version and compute node rpc version are not compatible. I try different solution to set the upgrade_levels on the newton controller an the compute node but nothing works.

Please can you give me some hints on how I can come out ahead.

Update: We have isolated the problem. The Issue depence on the servicegroup_driver. If we use mc as servicegroup driver it only works with newton controler -> newton node, if we use db as servicegroup driver it als works with newton controler -> mitaka node.