nova-scheduler doesnt start instances on new compute node (juno)

asked 2015-09-09 03:11:41 -0500

goran.tornqvist gravatar image

I got a allinone lab environment with Red Hat OpenStack 6.0 with external network working fine. I then added another compute node using packstack and all appears fine except that all instances that I launch using the nova availability zone is started on the primary node. If I launch an instance using the cli and specify the new host, or if I create a host aggregate in the Web GUI with only the new compute node in and launch the instance, it launches correctly.

Only once the primary host is maxed out of RAM and is almost swapping, it starts launcing instances on the other compute node.

My guess is that there is some nova-scheduler magic here that I dont understand. Any ideas whats going on?

My /var/log/nova/nova-scheduler.log says WARNING nova.scheduler.host_manager [req-d5fc782f-fb50-4b57-b66e-0ccc315ea87d None] Host has more disk space than database expected (159gb > 123gb) Everytime I launch an instance - not sure if it is related.

edit retag flag offensive close merge delete