We are currently using LINSTOR controllers in a HA configuration accessible from a floating IP managed by Corosync+Pacemaker.
Our cinder.conf uses the following configuration (organization specific data obfuscated) for the LINSTOR backend:
. . .
[KALMYK]
iscsi_helper = tgtadm
linstor_controller_diskless = False
linstor_default_resource_size = 1
linstor_default_storage_pool_name = KalmykLINSTOR
linstor_default_uri = linstor://XXX.XXX.XXX.6
linstor_default_volume_group = vg_kalmyk_linstor
linstor_volume_downsize_factor = 4096
volume_backend_name = LINSTOR
volume_driver = cinder.volume.drivers.linstordrv.LinstorDrbdDriver
. . .
However, the cinder-volume service ignores the linstor_default_uri specified and defaults to using linstor://localhost, errors-out when unable to do so, and goes into a crash-backoff loop.
We've identified the code responsible for this behavior and prepared a patch/diff file:
linstor.txt
After applying this patch/diff, the cinder-volume service accepts the specified URI and works properly in sending API requests to our controller.
Can these changes be verified?
We are currently using LINSTOR controllers in a HA configuration accessible from a floating IP managed by Corosync+Pacemaker.
Our
cinder.confuses the following configuration (organization specific data obfuscated) for the LINSTOR backend:However, the
cinder-volumeservice ignores thelinstor_default_urispecified and defaults to usinglinstor://localhost, errors-out when unable to do so, and goes into a crash-backoff loop.We've identified the code responsible for this behavior and prepared a patch/diff file:
linstor.txt
After applying this patch/diff, the
cinder-volumeservice accepts the specified URI and works properly in sending API requests to our controller.Can these changes be verified?