Hi Massimo,

If you are planning to use glance-direct import method then you need to configure os_glance_staging_store using a shared filesystem. If you plan to use only web-download and copy-image import methods then you don't need a shared filesystem for os_glance_staging_store.

If glance-direct needs to be used then the below path needs to be on a shared file system.
[os_glance_staging_store]
filesystem_store_datadir = /path/to/staging/store

If only web-download and copy-image import methods will be used then still you need to set the above section in glance-api.conf but no need for a shared filesystem in this case.


Thanks & Best Regards,

Abhishek Kekane


On Fri, Mar 26, 2021 at 11:29 AM Massimo Sgaravatto <massimo.sgaravatto@gmail.com> wrote:
We are using OpenStack Train but we plan to migrate soon to something newer

The glance-api service is deployed on 2 nodes.

We used to have 2 backends for glance (file and rbd) but now we want to use only ceph. So I guess the configuration for the stores can be simply something like this [*]

Reading:
https://docs.openstack.org/glance/train/admin/multistores.html

is not too clear to me if I have to configure somewhere filesystem_store_datadir.
In particular do I have to configure the  [os_glance_tasks_store] and
[os_glance_staging_store] reserved stores ?
If so, do I still need shared (between the 2 glance-api nodes )filesystems
for them ?

Thanks, Massimo


[*]

[DEFAULT]
enabled_backends = rbd:rbd

[glance_store]
default_backend = rbd


[rbd]
store_description = Ceph in replica 3
rbd_store_pool = glance-cloudtest
rbd_store_user = glance-cloudtest
rbd_store_ceph_conf = /etc/ceph/ceph.conf
rbd_store_chunk_size = 8