Read-only cinder volumes after cinder-volume container restart

Bug #1972168 reported by Matt Anson
12
This bug affects 2 people
Affects Status Importance Assigned to Milestone
kolla-ansible
In Progress
Undecided
Jakub Darmach

Bug Description

Hi,

On a Kolla-ansible deployed cloud using the Cinder NFS driver and a NetApp storage appliance, some cinder volumes become read-only when cinder-volume is restarted.
This appears to be caused by volume backing-file ownership in /var/lib/cinder.

I don't know if this is unique to the use of NFS exported from a NetApp appliance, but the NetApp NFS share mountpoint
is /var/lib/cinder/mnt/{UUID}, putting volume backing-files at /var/lib/cinder/mnt/{UUID}/volume-{UUID},
not directly under /var/lib/cinder.

This means that the recursive chown operations in both:
https://github.com/openstack/kolla-ansible/blob/ec4af6c90ee94e0f82f13afc09a9a7770bab165c/ansible/roles/cinder/templates/cinder-volume.json.j2#L39-L42
https://github.com/openstack/kolla/blob/b02f643db03694074293858470a1db36246b4a0e/docker/cinder/cinder-volume/extend_start.sh#L3-L5

Result in chown operations on the NFS mountpoint, rather than directly on the volume backing files as intended (I assume!).

Once the correct permissions are applied by hand to the read-only volume backing-files, normal service is resumed.

Revision history for this message
Jakub Darmach (darmachj) wrote :

Added the mounth path to exclude section in the config.json, this should help it.

Changed in kolla-ansible:
assignee: nobody → Jakub Darmach (darmachj)
Changed in kolla-ansible:
status: New → In Progress
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.