Support delete_on_termination in server attach volume

https://blueprints.launchpad.net/nova/+spec/support-delete-on-termination-in-server-attach-volume

This blueprint proposes to support passing delete_on_termination during volume attach so the attached volume can be deleted when the server is deleted.

Problem description

Currently, nova already supports the volume attach API, but it is not possible to configure whether the data volumes can be deleted when the instance is destroyed while the volume is being attached. This is a bit awkward when configuring the server to handle the data volume in the destroy instance.

Use Cases

In large scale environment, lots of resources were created in system, and sometimes an instance needs to be attached with more data volumes. Therefore, the user needs to set the processing mode of the attached volume for each instance. When destroying the instance, the data volume can be deleted together, the invalid data is cleared, and the storage space is released.

Proposed change

Add a new microversion to volume attach API to support configuring whether to delete the data volume when the instance is destroyed.

In the same microversion, add delete_on_termination to the GET responses when showing attached volumes.

See the REST API impact section for details.

Alternatives

The user cleans up the data volumes manually after deleting the server as they would have to do today.

Data model impact

None

REST API impact

URL: /v2.1/servers/{server_id}/os-volume_attachments

  • Request method: POST (attach volume)

    Add the delete_on_termination parameter to the request body with the same semantics/schema as the initial server create block_device_mapping_v2 object.

    {
  "volumeAttachment": {
    "volumeId": "a26887c6-c47b-4654-abb5-dfadf7d3f804",
    "delete_on_termination": true
  }
}

    The default value of the delete_on_termination field is False if not specified.

  • Request method: GET (list volume attachments)

    Add the delete_on_termination field to the response payload for attached volumes.

    {
  "volumeAttachments": [
    {
      "device": "/dev/sdd",
      "id": "a26887c6-c47b-4654-abb5-dfadf7d3f803",
      "serverId": "fb6077e6-c10d-4e81-87fa-cb0f8c103051",
      "tag": "foo",
      "volumeId": "a26887c6-c47b-4654-abb5-dfadf7d3f803",
      "delete_on_termination": false
    },
    {
      "device": "/dev/sdc",
      "id": "a26887c6-c47b-4654-abb5-dfadf7d3f804",
      "serverId": "fb6077e6-c10d-4e81-87fa-cb0f8c103051",
      "tag": null,
      "volumeId": "a26887c6-c47b-4654-abb5-dfadf7d3f804",
      "delete_on_termination": true
    }
  ]
}

URL: /servers/{server_id}/os-volume_attachments/{volume_id}

  • Request method: GET (show volume attachment)

    Add the delete_on_termination field to the response payload for attached volume.

    {
  "volumeAttachment": {
    "device": "/dev/sdc",
    "id": "a26887c6-c47b-4654-abb5-dfadf7d3f804",
    "serverId": "fb6077e6-c10d-4e81-87fa-cb0f8c103051",
    "tag": null,
    "volumeId": "a26887c6-c47b-4654-abb5-dfadf7d3f804",
    "delete_on_termination": true
  }
}

Note

PUT /servers/{server_id}/os-volume_attachments/{volume_id} is not part of this proposed change since that API today is only implemented for the “swap volume” operation which is only implemented by the libvirt driver. [1] Modifying the PUT API is out of scope for this spec. If a user wishes to change the delete_on_termination value of a non-root attached volume, they can do so by detaching and re-attaching the volume with the updated delete_on_termination value.

Security impact

None

Notifications impact

None

Other end user impact

python-novaclient and python-openstack client will need to be updated to support the new delete_on_termination parameter when attaching a volume and listing/showing attached volumes.

Performance Impact

None

Other deployer impact

None

Developer impact

None

Upgrade impact

Depending on implementation there should be no upgrade impact. Today when a volume is attached to a non-shelved-offloaded server, the BlockDeviceMapping record is created in the nova-compute service. When attaching a volume to a shelved offloaded server, the BDM is created in the API service. To avoid issues with trying to attach a volume with delete_on_termination=true to a server running on an older compute service, the implementation should just set the field in the API rather than the compute service.

Implementation

Assignee(s)

Primary assignee:

Brin Zhang

Work Items

  • Add delete_on_termination support in POST and GET os-volume_attachments APIs.

  • Add delete_on_termination support in python-novaclient and python-openstackclient.

  • Add related tests

Dependencies

None

Testing

  • Add related unit tests for negative scenarios such as trying to specify delete_on_termination during volume attach with an older microversion, passing delete_on_termination with an invalid value like null, etc.

  • Add related functional tests for normal scenarions, e.g. API samples.

Tempest testing should not be necessary since in-tree functional testing with the CinderFixture should be sufficient for testing this feature.

Documentation Impact

Update the API reference for the affected APIs.

References

History

Revisions

Release Name

Description

Train

Introduced