When creating the first OSD in a cluster, the method requested was being
ignored - meaning an LVM OSD would be created in all cases. If a given cluster
couldn't support LVM, it could not be deployed. If we relay the method value
requested by the cephadm invocation, we can create OSDs as expected.
Fixes: https://tracker.ceph.com/issues/72381
Signed-off-by: Zack Cerza <zack@cerza.org>
placement=PlacementSpec(host_pattern=host),
data_devices=DeviceSelection(paths=devices),
unmanaged=False,
+ method=drive_group.method,
objectstore=drive_group.objectstore
)