So i checked this with Abhishek Kekane today and we found that issue is related to ceph client(octopus) and os-brick. The issue happened when cinder(backend is rbd) is used as glance backend and ceph octopus client is used.
os-brick creates a temp configuration file[1] and uses it to connect to ceph cluster. With ceph octopus this do not work without [global] section defined for mon_host.
[global] <<- without this in temp configuration, ceph octopus client do not work, nautils client works fine
mon_host = <ip>:3300
octopus clients are coming as ceph master repo is used in kolla currently [2], a patch is proposed to use nautilus repo from Storage SIG https://review.opendev.org/#/c/707338/ in kolla. So atleast TripleO jobs shouldn't see this issue after that patch is merged and new container images are build from it.
And fixing(add [global] section while create config file) in os-brick could also be done as that should work with both nautilus and octopus ceph. Me don't know if this scenario is tested in glance/cinder in non TripleO job and if ceph octopus or ceph nautilus is used there. I think this is not being tested upstream otherwise the issue would have caught already or may be it's handled in different way.
So i checked this with Abhishek Kekane today and we found that issue is related to ceph client(octopus) and os-brick. The issue happened when cinder(backend is rbd) is used as glance backend and ceph octopus client is used.
os-brick creates a temp configuration file[1] and uses it to connect to ceph cluster. With ceph octopus this do not work without [global] section defined for mon_host.
[global] <<- without this in temp configuration, ceph octopus client do not work, nautils client works fine
mon_host = <ip>:3300
octopus clients are coming as ceph master repo is used in kolla currently [2], a patch is proposed to use nautilus repo from Storage SIG https:/ /review. opendev. org/#/c/ 707338/ in kolla. So atleast TripleO jobs shouldn't see this issue after that patch is merged and new container images are build from it.
And fixing(add [global] section while create config file) in os-brick could also be done as that should work with both nautilus and octopus ceph. Me don't know if this scenario is tested in glance/cinder in non TripleO job and if ceph octopus or ceph nautilus is used there. I think this is not being tested upstream otherwise the issue would have caught already or may be it's handled in different way.
[1] https:/ /github. com/openstack/ os-brick/ blob/a6dff156d7 0cd79e87c71bea5 3aa634311f22459 /os_brick/ initiator/ connectors/ rbd.py# L98-L99 /github. com/openstack/ kolla/blob/ master/ docker/ base/ceph_ master. repo
[2] https:/