I’ve deployed the Openstack base bundle for a charmed Openstack on Ubuntu 20.04 Focal using the yoga version. I’ve noticed that for some reason an LVM backend is deployed, however, that is not what I want. I want to use a custom storage array that I’ve already configured and is working fine. How can I disable the LVM cinder backend?
itclinux@scmaas01:~$ openstack volume service list
±-----------------±--------------------------±-----±--------±------±---------------------------+
| Binary | Host | Zone | Status | State | Updated At |
±-----------------±--------------------------±-----±--------±------±---------------------------+
| cinder-volume | controller02@LVM | nova | enabled | down | 2023-07-25T16:30:51.000000 |
| cinder-scheduler | controller01 | nova | enabled | down | 2023-07-25T16:30:59.000000 |
| cinder-volume | controller01@LVM | nova | enabled | down | 2023-07-25T16:30:57.000000 |
| cinder-scheduler | controller02 | nova | enabled | down | 2023-07-25T16:30:53.000000 |
| cinder-volume | controller03@LVM | nova | enabled | down | 2023-07-25T16:30:51.000000 |
| cinder-scheduler | controller03 | nova | enabled | down | 2023-07-25T16:30:53.000000 |
| cinder-volume | cinder@cinder-array | nova | enabled | up | 2023-07-26T15:28:25.000000 |
| cinder-scheduler | cinder | nova | enabled | up | 2023-07-26T15:28:27.000000 |
±-----------------±--------------------------±-----±--------±------±---------------------------+
no it isn’t Those nodes are actually up and running.
can you share the output of “juju status”?
` Sure, this is the link to the output. https://pastes.io/j7ai82kzn6. As an Important note, the Ceph configuration is only linked to glance with ceph-fs-gw to provide storage space for images. I’m using a storage configuration.
did you have LVM configured at some and you just reconfigured your system later?
No I didn’t. I deployed the cloud without a storage backend, then deployed my custom charm and did the cinder relation with that custom charm.
Does your custom charm relate to cinder over the storage-backend interface?. The output of juju status --relations cinder cinder-hitachie590 could give us some insights.
Yes, it does relate over the storage-backend interface.
if the custom charm was deployed after (not part of the bundle), then that would be explanation why the LVM backend was enabled, and then disabled when the relation was added.
you can add your custom charm to a bundle with something like this:
How can I deploy my bundle without any backend or LVM so I can later on deploy my custom charm?
based on the source code I pointed out, this is not possible, the comment left in that section this is because otherwise the service crashes and gets restarted over and over again. This could be considered as a bug, feel free to file one at https://bugs.launchpad.net/charm-cinder/+filebug
I didn’t know that was a possibility. I thought that only Canonical Charmers could do that, but I’m interested on the possibility and learning more about how charms work so if you could point me in the right direction that would be perfect.
Right now the charm shows as waiting, even though everything is working and the /etc/cinder/cinder.conf file has been properly configured.
cinder/0* active idle 0 xxx.xxx.xxx.103 8776/tcp Unit is ready
cinder-hacluster/0* active idle xxx.xxx.xxx.103 Unit is ready and clustered
cinder-hitachie590/0* waiting idle xxx.xxx.xxx.103 Charm configuration in progress
cinder-mysql-router/0* active idle xxx.xxx.xxx.103 Unit is ready
You can implement a handler for the ‘update-status’ event , the handler needs to verify if everything is OK (e.g. is my payload service running?, is there a service listening in a tcp port I manage?, etc), see https://juju.is/docs/sdk/update-status-event
