If you need to perform maintenance on a host, there are some important steps to take beforehand:
-If the host is a compute node:
- Live-migrate all VMs off that host:
- Get list of VMs and their IDs: nova list --host <node name> --all-tenants
- Migrate each off: nova live-migration <vm id>
- Run a. again to make sure it was migrated
- Disable new VMs from being scheduled (must be run from a nova management node): nova-manage service disable --host=<node name> --service=nova-compute
Create maintenance window in Nagios so that alerts aren't generated and people don't panic.
-If the host is a Ceph Storage Node:
It is crucial that not more than one Ceph node is brought down at once. After restarting, ensure that the rebalance is complete by running ceph -s and ensuring a HEALTH OK status.
-If the host is a admin / control plane hypervisor: