Cinder
======

.. literalinclude:: cinder.yaml

Cinder using Netapp
-------------------

.. literalinclude:: cinder_netapp.yaml

The configmap that contains the netapp copyoffload tool can be created using this command: 
``kubectl create configmap netapp-copyoffload --from-file na_copyoffload_64``


Cinder volume active-active high-availability
---------------------------------------------

Setting ``cluster: true`` for a volume backend will enable active-active cluster.
Switching to cluster or back will **not** migrate volumes resources. With enabled
cluster mode, replicas can be increased above one. This is not possible for
non-cluster setups where the yaook operator will throw an error.
Do not set ``backend_host`` config in cluster mode except for single replicas.

Currently netapp-nfs driver (2023.2 and above) and ceph is supported in yaook.


Convert an existing service to a cluster
----------------------------------------

The cinder-volume service includes an automatic volume migration if it is started with
the same host (or backend_host if set) and new cluster configuration. So having
already set a ``backend_host`` is the requirement for this to work.

To trigger the auto-migration keep ``backend_host`` config, set cluster to true and
keep replicas at one. After the StatefulSet comes up the pod should log migrated volumes.

If not, perform the following manual steps:

1. Access the `cinder-api` container:

   .. code-block:: bash

      yaookctl shell cinder api

2. Update all volumes to point to the new clustered host (replace hostnames accordingly):

   .. code-block:: bash

      cinder-manage volume update_host --current <old_host> --new <new_host>

3. Update the `service_uuid` for all volumes.
   This step is required due to upstream Cinder bugs:

   -  `Cinder Bug #1890278: cinder-manage volume update_host does not update
      service uuids <https://bugs.launchpad.net/cinder/+bug/1890278>`__
   -  `Cinder Bug #2117273: volume service_uuid not updated after
      migration <https://bugs.launchpad.net/cinder/+bug/2117273>`__

   .. code-block:: bash

      cinder-manage volume update_service

Finally remove ``backend_host`` configuration and scale up replicas.

An unused cinder-volume service for the cluster will remain but can safely be
removed with ``yaookctl shell cinder api cinder-manage service <list|remove>``.