On 6/17/21 07:24, Ramesh Ramanathan B wrote:
Dear All,
One observation we have while using Open Stack Rocky is, when a compute node goes down, the VM status still shows active (the servers running in the compute node that went down). Is this the expected behavior? Any configurations required to get the right status.
In the attached image the compute is down, but the VM status still shows active. We are running a data center so it is not practical to run nova reset-state for all the servers. Is there an API to force update Nova to show the correct status? Or any configurations missing that is causing this?
This is expected behavior. The way we expose this condition is via the host_status field on the VM. By default, this is exposed only via the admin API but can be changed as desired via policy [1][2]. For example (this is admin):
$ nova list --fields id,name,status,task_state,power_state,host_status,networks +--------------------------------------+-------+---------+------------+-------------+-------------+--------------------------------------+ | ID | Name | Status | Task State | Power State | Host Status | Networks | +--------------------------------------+-------+---------+------------+-------------+-------------+--------------------------------------+ | f1aece4e-e7f7-436b-9faf-799129503dcc | eight | SHUTOFF | None | Shutdown | UP | public=2001:db8::3aa, 192.168.33.189 | +--------------------------------------+-------+---------+------------+-------------+-------------+--------------------------------------+
The initial proposal was to make the VM status UNKNOWN when the host is down but the consensus was to stick with the host_status and keep it separate from the VM status. See the abandoned spec for details [3]. HTH, -melwitt [1] os_compute_api:servers:show:host_status:unknown-only in https://docs.openstack.org/nova/latest/configuration/policy.html [2] https://blueprints.launchpad.net/nova/+spec/policy-rule-for-host-status-unkn... [3] https://review.opendev.org/c/openstack/nova-specs/+/666181