Storage: Use reverter for Ceph RBD volume refresh #13198
Conversation
|
Mhh, I want to check one more thing before we merge. The generic methods probably should return a |
…Volume Signed-off-by: Julian Pelizäus <julian.pelizaeus@canonical.com>
roosterfish
force-pushed
the
rbd_refresh_revert
branch
from
8622442 to
817a077
Compare
This adds a new refreshVolume function which allows returning cleanup hooks. This is useful as a VMs filesystem volume gets refreshed using the generic approach whilst block volumes use the optimized way of refreshing RBD backed volumes. Signed-off-by: Julian Pelizäus <julian.pelizaeus@canonical.com>
Signed-off-by: Julian Pelizäus <julian.pelizaeus@canonical.com>
Signed-off-by: Julian Pelizäus <julian.pelizaeus@canonical.com>
…olume Signed-off-by: Julian Pelizäus <julian.pelizaeus@canonical.com>
Signed-off-by: Julian Pelizäus <julian.pelizaeus@canonical.com>
…teVolumeFromMigration Signed-off-by: Julian Pelizäus <julian.pelizaeus@canonical.com>
Signed-off-by: Julian Pelizäus <julian.pelizaeus@canonical.com>
…umeFromMigration Signed-off-by: Julian Pelizäus <julian.pelizaeus@canonical.com>
…eFromMigration Signed-off-by: Julian Pelizäus <julian.pelizaeus@canonical.com>
…eFromMigration Signed-off-by: Julian Pelizäus <julian.pelizaeus@canonical.com>
…eVolumeFromMigration Signed-off-by: Julian Pelizäus <julian.pelizaeus@canonical.com>
…eFromMigration Signed-off-by: Julian Pelizäus <julian.pelizaeus@canonical.com>
roosterfish
force-pushed
the
rbd_refresh_revert
branch
from
817a077 to
084e441
Compare
|
Ok so this is now ready for review again. In order to properly cleanup the volumes that get refreshed using the generic approach, the respective functions now return the I have moved most of the logic into unexported equivalents of the driver functions so that those can be called twice but return the cleanup hooks for each of their invocations. The actual drivers function can now collect the cleanups and react appropriately in case of error. |
This ensures snapshots are cleaned up in case the refresh fails. Otherwise this leads to error messages for subsequent refreshes as some of the snapshots already exist.