I’m trying to create a custom storage backend to connect to my storage array. I’ve used the cookiecutter template that should be used to develop custom backends. here
Can you test changing ‘yoga’ to ‘quincy’ as the release name ? Additionally, you may try changing the call from get_charm_class_for_release into get_charm_class (without modifying the release name).
my configuration doesn’t work. I need the compute nodes to behave like they are storage nodes and I need my cinder-storage-node services to run on the compute nodes for them to get correctly configured. Look at this service list:
+------------------+-------------------------------+------+---------+-------+----------------------------+
| Binary | Host | Zone | Status | State | Updated At |
+------------------+-------------------------------+------+---------+-------+----------------------------+
| cinder-volume | compute03@VSPG350 | nova | enabled | up | 2022-09-28T16:27:10.000000 |
| cinder-scheduler | controller | nova | enabled | up | 2022-09-28T16:27:10.000000 |
| cinder-volume | compute04@VSPG350 | nova | enabled | up | 2022-09-28T16:27:03.000000 |
| cinder-volume | compute01@VSPG350 | nova | enabled | up | 2022-09-28T16:27:11.000000 |
| cinder-volume | compute05@VSPG350 | nova | enabled | up | 2022-09-28T16:27:07.000000 |
| cinder-volume | compute02@VSPG350 | nova | enabled | up | 2022-09-28T16:27:06.000000 |
| cinder-volume | compute06@VSPG350 | nova | enabled | up | 2022-09-28T16:27:03.000000 |
+------------------+-------------------------------+------+---------+-------+----------------------------+
I’m not entirely sure to be honest. When you mention that the configuration “doesn’t work”, what do you mean exactly ? Is the configuration propagated correctly to the cinder units ?
I think what you’re trying to achieve could be doable at deploy time via juju, although I need more details as to what isn’t working.
What I mean is that everything from Canonical point of view works just fine. The real issue is that my cinder driver is deployed and configured to use FC interfaces and to plug in volumes via FC to the HBA Ports on the compute nodes.
As you can see, the control nodes can run cinder-scheduler, and cinder-api service without issues, but the cinder-volume service needs to run on the compute nodes, since the cinder-volume service is the one that connects to the storage and then connects the LUN to the host. Is it possible to achieve that using juju? I can do it by hand, so I’m guessing it should be doable.