I don't have an answer. Just another variation of the same problem.

• Created a ceph volume using boot from image and create volume in the horizone dashboard
• It fails and leaves a volume that does not match any ceph rbd
• When I force-delete the volume it does nothing
• When I delete the volume it says : "ERROR: Invalid volume: Volume status must be available or error, but current status is: error_deleting"

How can I get out of there without tampering with the database ?

I tried following directly from mysql and worked for me

Connect to your compute node instance with SSH client(I use putty)

> cinder list

> cinder snapshot-list

More cinder command can be found here http://docs.openstack.org/cli-reference/content/cinderclient_commands.html (http://docs.openstack.org/cli-referen...)

Note down the snapshot id and volume id that you want to remove

now connect to mysql server cinder database and delete entry

> mysql -u root cinder

mysql> delete from snapshots where id='xxxxxxx-xxx-xxx-xxx-xxxxxxxxxx'

mysql> delete from volumes where id='xxxxxxx-xxx-xxx-xxx-xxxxxxxxxx'

done.

Now snapshot and volume both removed from Volumes & Snapshots listing.

Make sure you delete snapshot first before remove volume.

Volume should NOT be deleted before connected snapshots are deleted.

hi

A similar problem I encountered too. I have cinder configured to use a ceph cluster as backend. I tried to create a volume form an image but because in cinder.conf I didn't have glance_host ip configured it tried to connect to glance to the default value (localhost). So the volume was not created and has a state 'error'.

I have reset the state but when I try to delete the image image the message /var/log/upstart/cinder-volume.log is: error opening image volume-xxxxxxxxxxxxxxxxxxxxx at snapshot None.

With the 'rbd -p volumes ls' I cannot see a volume corresponding to that id. Did anyone encounter a problem like that? Is is a more elegant solution than to hack the database?

I also got this problem. It turns out the volume was in use by the tgtd, so I did the following:

tgt-admin -s | grep <lv_id>

followed by

tgt-admin -delete iqn...openstack:<lv_id> (basically, what you get from grep).

Then I just force-delete-ed the volume to remove it from the cinder db.

I am using CentOS 7 with Ceph and ran into the same problem. I found that there were no /var/lock/cinder and /var/lock/nova directories on the compute/cinder nodes. I created them, gave them user and group permissions, and restarted services. After that I was able to delete stuck volumes and instances.

For stuck nova instances:

# mkdir /var/lock/nova
# chown -R nova:nova /var/lock/nova
# systemctl restart openstack-nova-compute.service

For stuck ceph volumes:

# mkdir /var/lock/cinder
# chown -R cinder:cinder /var/lock/cinder
# systemctl restart openstack-cinder-volume.service

Hope that helps.

Before going into the database and tweaking entries there you can check the /var/log/cinder/volume log to see if volume in question is reported as 'in use'. If yes you can follow the procedure here https://ask.openstack.org/en/question/87725/failed-to-delete-a-volume-as-the-backend-lv-is-shown-open/ (https://ask.openstack.org/en/question...) which worked really well for me.

HTH.

FYI, there is some work going on here https://review.openstack.org/#/c/184537/ to add a force volume detach command to nova-manage. Input on that review would be helpful for people having this issue.

yes same problem here also happend with ceph my problem was that i upgraded to havana and i forgot to set the approiate settings

so what happend was cinder-volume was down cinder-api was up cinder-scheduler was up

created a snapshot but volume was never created in ceph so it kept in creating status. then i tried to delete it no luck because i needed to change te state so after that i deleted it and then it was stuck in error_deleting because it could not find the image in ceph