No, I don't have these tools enabled to check the placement allocation and resources. I can try to. Meanwhile is there some configuration to be set in stein release as opposed to rocky towards cells or placement ? On the same nodes rocky was installed and it worked fine straight out of box. Stein is showing issues with controllers and says "Failed to compute_task_build_instances: No valid host was found." under nova-conductor.log. Also under nova-scheduler.log I could see a single info entry: " Got no allocation candidates from the Placement API. This could be due to insufficient resources or a temporary occurence as compute nodes start up." But I do see that all compute nodes are enabled and is up. On Mon, Feb 8, 2021, 4:03 PM Pierre Riteau <pierre@stackhpc.com> wrote:
You can also use the osc-placement CLI to check what the placement data looks like. Do you have resource providers and matching inventories for your hypervisors?
On Mon, 8 Feb 2021 at 11:16, roshan anvekar <roshananvekar@gmail.com> wrote:
VM creation is not getting scheduled itself showing no valid host found.
The logs show no error in scheduler logs too
On Mon, Feb 8, 2021, 2:44 PM Mark Goddard <mark@stackhpc.com> wrote:
On Fri, 5 Feb 2021 at 15:49, roshan anvekar <roshananvekar@gmail.com>
Thanks for the reply.
Well, I have a multinode setup ( 3 controllers and multiple compute
nodes) which was initially deployed with rocky and was working fine.
I checked the globals.yml and site.yml files between rocky and stein
and I could not see any significant changes.
Also under Admin-Compute-Hypervisor, I see that all the compute nodes
are showing up under Compute section. the hypervisor section is empty.
I was wondering if controllers are placed under a different aggregate
and not able to show up. I can see all 3 controllers listed in host-aggregates panel though and are in service up state.
VM creation fails with no valid host found error.
I am not able to point out issue since I don't see any errors in
deployment too.
Regards, Roshan
On Fri, Feb 5, 2021, 7:47 PM Sean Mooney <smooney@redhat.com> wrote:
On Fri, 2021-02-05 at 14:53 +0530, roshan anvekar wrote:
Hi all,
Scenario: I have an installation of Openstack stein through
kolla-ansible.
The deployment went fine and all services look good.
Although I am seeing that under Admin--> Compute --> Hypervisors
horizon, all the controller nodes are missing. It's a blank list. did you actully deploy the nova compute agent service to them?
that view is showing the list of host that are running the nova compute service typically that is not deployed to the contolers.
host in the contol group in the kolla multi node inventlry
https://github.com/openstack/kolla-ansible/blob/master/ansible/inventory/mul...
are not use to run the compute agent by default only nodes in the compute group are
https://github.com/openstack/kolla-ansible/blob/master/ansible/inventory/mul...
the eception to that is ironic https://github.com/openstack/kolla-ansible/blob/master/ansible/inventory/mul... which is deployed to the contolers.
the nova compute agent used for libvirt is deployed specificlly to
wrote: panel in the compute hosts via the nova-cell role at least on master
https://github.com/openstack/kolla-ansible/blob/master/ansible/nova.yml#L118
this was done a little simpler before adding cell support but the inventory side has not changed in many release in this regard.
Also "Openstack hypervisor list" gives an empty list.
I skimmed through the logs and found no error message other than in nova-scheduler that:
*Got no allocation candidates from the Placement API. This could
be due to
insufficient resources or a temporary occurence as compute nodes start up.*
Subsequently I checked placement container logs and found no error message or anamoly.
Not sure what the issue is. Any help in the above case would be appreciated.
Regards, Roshan Nova compute logs are probably the best place to go digging.