Thank you for the clarification. So in order to deploy glance on multiple nodes, I need to first set up an NFS storage then specify the shared path either by overriding the content of ansible/group_vars/all.yml or by adding glance_file_datadir_volume:NFS_PATH in globals.yml ? Which NFS tool do you think I should use? Thank you again. Regards On Thu, 21 Jul 2022 at 10:42, Pierre Riteau <pierre@stackhpc.com> wrote:

With the default backend (file), Glance is deployed on a single controller, because it uses a local Docker volume to store Glance images. This is explained in the documentation [1]: "By default when using file backend only one glance-api container can be running". See also the definition of glance_api_hosts in ansible/group_vars/all.yml.

If you set glance_file_datadir_volume to a non-default path, it is assumed to be on shared storage and kolla-ansible will automatically use all glance-api group members.

You can also switch to another backend such as Ceph or Swift.

[1] https://docs.openstack.org/kolla-ansible/latest/reference/shared-services/gl...

On Thu, 21 Jul 2022 at 10:57, A Monster <amonster369@gmail.com> wrote:

...

I've deployed openstack xena using kolla ansible on a centos 8 stream cluster, using two controller nodes, however I found out after the deployment that glance api is not available in one node, I tried redeploying but I got the same behavior, although the deployment finished without displaying any error.

Thank you. Regards

On Mon, 2022-12-19 at 22:29 +0530, pradeep wrote:

Hi,

Sorry to hijack this thread. I have a similar issue. I have used netapp NAS share, mounted in all the 3 controllers as /glanceimages and modified glance_file_datadir_volume:/glanceimages in globals.yml. I have tried deploy and reconfigure commands but was unlucky. It doesn't deploy glance-api containers in controller2 and 3. Please let me know if you succeed in this. looking at the glance-api group it is defiend as the childeren of the glance group https://github.com/openstack/kolla-ansible/blob/master/ansible/inventory/mul... [glance-api:children] glance

which in trun is defiend as the childern of the controler group https://github.com/openstack/kolla-ansible/blob/master/ansible/inventory/mul... [glance:children] control looking breifly at the glance role the api will be deployed if its enabled and the current host is in the relevent group https://github.com/openstack/kolla-ansible/blob/master/ansible/roles/glance/... host_in_group is defiend here https://github.com/openstack/kolla-ansible/blob/master/ansible/roles/glance/... host_in_groups: "{{ inventory_hostname in glance_api_hosts }}" glance_api_hosts is defined here https://github.com/openstack/kolla-ansible/blob/ae3de342e48bc4293564a3b532a6... glance_api_hosts: "{{ [groups['glance-api'] | first] if glance_backend_file | bool and glance_file_datadir_volume == 'glance' else groups['glance- api'] }}" so that is what is making it only work with the first host. with that said you have set glance_file_datadir_volume:/glanceimages in the globals.yaml so that should have also caused the else banch to be taken. i have not worked on kolla-ansible for 5+ years at this point but it looks like the ablity to use multiple api instance was modifed by https://github.com/openstack/kolla-ansible/commit/d57c7019a92f498f8dd876f4da... for https://bugs.launchpad.net/kolla-ansible/+bug/1722422 but that simply moved where that variable was defined the actual behavior change was intoduced by https://github.com/openstack/kolla-ansible/commit/4fde486dc8b29b8d087ab4bfff... for https://bugs.launchpad.net/kolla-ansible/+bug/1681301 the intent of that change was not to prevent you runnign multipel glance api servers but just to ensure that by default only one instance was deployed in a multinode config. global.yaml is passed with -e to ansible by kolla-ansible so defining "glance_file_datadir_volume:/glanceimages" should have been enough to enable multiple glance instnaces to be deployed. so it sounds like you did everything correctly but perhaps the kolla-ansible team can point you in the right direction. at first glance you could try defining  glance_api_hosts: [groups['glance-api']] in the global.yaml but that should not be required if you have glance_file_datadir_volume defined to a non default value.

Regards Pradeep

On Mon, 19 Dec 2022 at 14:49, A Monster <amonster369@gmail.com> wrote:

...

Thank you for the clarification. So in order to deploy glance on multiple nodes, I need to first set up an NFS storage then specify the shared path either by overriding the content of ansible/group_vars/all.yml or by adding glance_file_datadir_volume:NFS_PATH in globals.yml ? Which NFS tool do you think I should use? Thank you again. Regards

On Thu, 21 Jul 2022 at 10:42, Pierre Riteau <pierre@stackhpc.com> wrote:

...

With the default backend (file), Glance is deployed on a single controller, because it uses a local Docker volume to store Glance images. This is explained in the documentation [1]: "By default when using file backend only one glance-api container can be running". See also the definition of glance_api_hosts in ansible/group_vars/all.yml.

If you set glance_file_datadir_volume to a non-default path, it is assumed to be on shared storage and kolla-ansible will automatically use all glance-api group members.

You can also switch to another backend such as Ceph or Swift.

[1] https://docs.openstack.org/kolla-ansible/latest/reference/shared-services/gl...

On Thu, 21 Jul 2022 at 10:57, A Monster <amonster369@gmail.com> wrote:

...

I've deployed openstack xena using kolla ansible on a centos 8 stream cluster, using two controller nodes, however I found out after the deployment that glance api is not available in one node, I tried redeploying but I got the same behavior, although the deployment finished without displaying any error.

Thank you. Regards