cinder — Scenarios

sequenceDiagram
  autonumber
  actor U as User
  participant API as cinder-api
  participant DB as MariaDB
  participant MQ as RabbitMQ
  participant SCH as cinder-scheduler
  participant VOL as cinder-volume (host@backend)
  participant G as Glance
  participant ARR as Storage array
  U->>API: POST /v3/{project_id}/volumes {"volume": {"size": 10, "imageRef": "…"}}
  API->>G: GET /v2/images/{id} (size, min_disk, status)
  API->>DB: SELECT quota_usages … FOR UPDATE; INSERT reservations
  alt quota exceeded
    API-->>U: 413 overLimit VolumeSizeExceedsAvailableQuota
  end
  API->>DB: INSERT volumes (status=creating); commit reservations
  API->>MQ: cast create_volume (topic cinder-scheduler)
  API-->>U: 202 {"volume": {"status": "creating", …}}
  MQ->>SCH: create_volume
  SCH->>SCH: filters (AZ, capacity, capabilities) + weighers on cached pool stats
  alt no valid backend
    SCH->>DB: UPDATE volumes SET status='error'; INSERT messages (VOLUME_VOLUME_001_003)
  else backend found
    SCH->>DB: UPDATE volumes SET host='node1@lvm#pool'
    SCH->>MQ: cast create_volume (topic cinder-volume.node1@lvm, server node1)
    MQ->>VOL: create_volume
    VOL->>VOL: create flow (notification volume.create.start)
    alt image already in the image-volume cache
      VOL->>ARR: clone the cached volume
    else cache miss
      VOL->>ARR: create volume
      VOL->>G: GET /v2/images/{id}/file
      VOL->>ARR: write the image (qemu-img convert through os-brick)
    end
    alt driver failure and attempts left
      VOL->>MQ: cast create_volume to the scheduler (retry, backend excluded)
    else success
      VOL->>DB: UPDATE volumes SET status='available', bootable=true; INSERT volume_glance_metadata
      VOL->>MQ: notify volume.create.end
    end
  end
  U->>API: GET /v3/{project_id}/volumes/{id} (polling)
  API-->>U: 200 status available | error

sequenceDiagram
  autonumber
  participant N as nova-compute
  participant API as cinder-api
  participant DB as MariaDB
  participant VOL as cinder-volume
  participant ARR as Storage array
  N->>API: POST /v3/{project_id}/attachments {"attachment": {"volume_uuid", "instance_uuid"}} (3.44+)
  API->>DB: UPDATE volumes SET status='reserved' WHERE status IN ('available','downloading') (or in-use if multiattach)
  API->>DB: INSERT volume_attachment (attach_status=reserved)
  API-->>N: 200 {"attachment": {"id", "status": "reserved", "connection_info": {}}}
  N->>API: PUT /v3/{project_id}/attachments/{id} {"attachment": {"connector": {"host", "initiator", "ip", "multipath"…}}}
  API->>API: lock cinder-attachment_update-<volume>-<host>
  API->>VOL: RPC call attachment_update(connector)
  VOL->>ARR: create_export + initialize_connection (map LUN to initiator)
  VOL->>DB: UPDATE volume_attachment SET connection_info, connector, attach_status='attaching'
  VOL-->>API: connection_info {"driver_volume_type": "iscsi", "data": {…}}
  alt driver error or RPC timeout
    API-->>N: 500 "Unable to update the attachment" / computeFault
  end
  API-->>N: 200 {"attachment": {"connection_info": {…}}}
  N->>N: os-brick connect_volume on the compute host
  N->>API: POST /v3/{project_id}/attachments/{id}/action {"os-complete": null}
  API->>DB: attachment attached; volume in-use, attach_status attached
  API-->>N: 204
  Note over N,API: Detach: DELETE /v3/{project_id}/attachments/{id}, sent with a service token, then RPC call attachment_delete (terminate_connection, remove_export)

sequenceDiagram
  autonumber
  actor U as User
  participant API as cinder-api
  participant DB as MariaDB
  participant SCH as cinder-scheduler
  participant VOL as cinder-volume
  participant NOVA as Nova API
  U->>API: POST /v3/{project_id}/volumes/{id}/action {"os-extend": {"new_size": 20}} (OpenStack-API-Version: volume 3.42)
  API->>DB: UPDATE volumes SET status='extending', previous_status='in-use' WHERE status='in-use'
  API->>DB: reserve quota gigabytes +10
  alt over quota
    API->>DB: status back to in-use
    API-->>U: 413 VolumeSizeExceedsAvailableQuota
  end
  API->>SCH: cast extend_volume
  API-->>U: 202
  SCH->>SCH: backend_passes_filters (capacity of the volume's own pool)
  alt not enough capacity
    SCH->>DB: status back to in-use; rollback quota; user message
  else
    SCH->>VOL: cast extend_volume
    VOL->>VOL: driver.extend_volume
    alt driver failure
      VOL->>DB: status error_extending; rollback quota; message VOLUME_VOLUME_007_010
    else
      VOL->>DB: commit quota; size=20; status in-use; notify volume.resize.end
      VOL->>NOVA: POST /v2.1/os-server-external-events [{"name": "volume-extended", "server_uuid", "tag": volume_id}] (2.51, [nova] service user)
      alt Nova error
        VOL->>DB: user message VOLUME_VOLUME_007_009
      end
    end
  end

sequenceDiagram
  autonumber
  actor U as User
  participant API as cinder-api
  participant DB as MariaDB
  participant SCH as cinder-scheduler
  participant BK as cinder-backup
  participant VOL as cinder-volume
  participant S as Backup store (Swift…)
  U->>API: POST /v3/{project_id}/backups {"backup": {"volume_id", "incremental": false, "force": true}}
  API->>DB: reserve quota backups, backup_gigabytes
  API->>DB: UPDATE volumes SET status='backing-up', previous_status=<status>
  API->>DB: INSERT backups (status creating)
  API->>SCH: cast create_backup
  API-->>U: 202 {"backup": {"id", "name"}}
  SCH->>SCH: pick a live cinder-backup in the AZ
  alt no backup service
    SCH->>DB: backup error, volume previous status, message BACKUP_SCHEDULE_ERROR
  else
    SCH->>BK: cast create_backup (topic cinder-backup.<host>)
    BK->>VOL: RPC call get_backup_device (temporary snapshot/clone if in-use)
    BK->>BK: attach the device locally (os-brick)
    loop each chunk
      BK->>S: PUT object (compressed, sha256 per chunk)
      BK-->>BK: notify backup.createprogress
    end
    BK->>DB: backup available (size, object count); volume back to previous status
  end
  U->>API: POST /v3/{project_id}/backups/{id}/restore {"restore": {"volume_id": …}}
  API->>DB: backup restoring; volume restoring-backup
  API->>BK: cast restore_backup
  BK->>S: GET objects
  BK->>DB: backup available; volume available (or error_restoring)

sequenceDiagram
  autonumber
  actor U as User
  participant API as cinder-api
  participant DB as MariaDB
  participant KM as Barbican
  participant VOL as cinder-volume
  participant ARR as Storage array
  U->>API: DELETE /v3/{project_id}/volumes/{id}
  API->>DB: conditional UPDATE status='deleting' (not attached, no snapshots, status in available/error/…)
  alt condition not met
    API-->>U: 400 InvalidVolume "Volume … must not be migrating, attached, …"
  end
  API->>DB: evict image-volume cache entry if any
  API->>KM: DELETE secret encryption_key_id
  alt key deletion fails
    API->>DB: status error_deleting
    API-->>U: 400 "Unable to delete encryption key for volume"
  end
  API->>VOL: cast delete_volume
  API-->>U: 202
  VOL->>VOL: lock cinder-<id>-delete_volume
  VOL->>ARR: remove_export, delete_volume
  alt array says busy
    VOL->>DB: status available
  else failure
    VOL->>DB: status error_deleting
  else success
    VOL->>DB: release quotas; soft-delete volume; notify volume.delete.end
  end