Share Replication¶

As of the Mitaka release of OpenStack, manila supports replication of shares between different pools for drivers that operate with driver_handles_share_servers=False mode. These pools may be on different backends or within the same backend. This feature can be used as a disaster recovery solution or as a load sharing mirroring solution depending upon the replication style chosen, the capability of the driver and the configuration of backends.

This feature assumes and relies on the fact that share drivers will be responsible for communicating with ALL storage controllers necessary to achieve any replication tasks, even if that involves sending commands to other storage controllers in other Availability Zones (or AZs).

End users would be able to create and manage their replicas, alongside their shares and snapshots.

Storage availability zones and replication domains¶

Replication is supported within the same availability zone, but in an ideal solution, an Availability Zone should be perceived as a single failure domain. So this feature provides the most value in an inter-AZ replication use case.

The replication_domain option is a backend specific StrOpt option to be used within manila.conf. The value can be any ASCII string. Two backends that can replicate between each other would have the same replication_domain. This comes from the premise that manila expects Share Replication to be performed between backends that have similar characteristics.

When scheduling new replicas, the scheduler takes into account the replication_domain option to match similar backends. It also ensures that only one replica can be scheduled per pool. When backends report multiple pools, manila would allow for replication between two pools on the same backend.

The replication_domain option is meant to be used in conjunction with the storage_availability_zone (or back end specific backend_availability_zone) option to utilize this solution for Data Protection/Disaster Recovery.

Replication types¶

When creating a share that is meant to have replicas in the future, the user will use a share_type with an extra_spec, replication_type set to a valid replication type that manila supports. Drivers must report the replication type that they support as the replication_type capability during the _update_share_stats() call.

Three types of replication are currently supported:

writable: Synchronously replicated shares where all replicas are writable. Promotion is not supported and not needed.
readable: Mirror-style replication with a primary (writable) copy and one or more secondary (read-only) copies which can become writable after a promotion.
dr (for Disaster Recovery): Generalized replication with secondary copies that are inaccessible until they are promoted to become the active replica.

Note

The term active replica refers to the primary share. In writable style of replication, all replicas are active, and there could be no distinction of a primary share. In readable and dr styles of replication, a secondary replica may be referred to as passive, non-active or simply replica.

Promotion¶

For readable and dr styles, we refer to the task of switching a non-active replica with the active replica as promotion. For the writable style of replication, promotion does not make sense since all replicas are active (or writable) at all given points of time.

The status attribute of the non-active replica being promoted will be set to replication_change during its promotion. This has been classified as a busy state and hence API interactions with the share are restricted while one of its replicas is in this state.

Promotion of replicas with replica_state set to error may not be fully supported by the backend. However, manila allows the action as an administrator feature and such an attempt may be honored by backends if possible.

When multiple replicas exist, multiple replication relationships between shares may need to be redefined at the backend during the promotion operation. If the driver fails at this stage, the replicas may be left in an inconsistent state. The share manager will set all replicas to have the status attribute set to error. Recovery from this state would require administrator intervention.

Snapshots¶

If the driver supports snapshots, the replication of a snapshot is expected to be initiated simultaneously with the creation of the snapshot on the active replica. Manila tracks snapshots across replicas as separate snapshot instances. The aggregate snapshot object itself will be in creating state until it is available across all of the share’s replicas that have their replica_state attribute set to active or in_sync.

Therefore, for a driver that supports snapshots, the definition of being in_sync with the primary is not only that data is ensured (within the recovery point objective), but also that any ‘available’ snapshots on the primary are ensured on the replica as well. If the snapshots cannot be ensured, the replica_state must be reported to manila as being out_of_sync until the snapshots have been replicated.

When a snapshot instance has its status attribute set to creating or deleting, manila will poll the respective drivers for a status update. As described earlier, the parent snapshot itself will be available only when its instances across the active and in_sync replicas of the share are available. The polling interval will be the same as replica_state_update_interval.

Access Rules¶

Access rules are not meant to be different across the replicas of the share. Manila expects drivers to handle these access rules effectively depending on the style of replication supported. For example, the dr style of replication does mean that the non-active replicas are inaccessible, so if read-write rules are expected, then the rules should be applied on the active replica only. Similarly, drivers that support readable replication type should apply any read-write rules as read-only for the non-active replicas.

Drivers will receive all the access rules in create_replica, delete_replica and update_replica_state calls and have ample opportunity to reconcile these rules effectively across replicas.

Understanding Replication Workflows¶

Creating a replica¶

The user has to specify the share name/id of the share that is supposed to be replicated and optionally an availability zone for the replica to exist in. The replica inherits the parent share’s share_type and associated extra_specs. Scheduling of the replica is similar to that of the share.

The ShareReplicationFilter will ensure that the pool is within
the same replication_domain as the active replica and also ensures that the pool does not already have a replica for that share.

Drivers supporting writable style must set the replica_state attribute to active when the replica has been created and is available.

Deleting a replica¶

Users can remove replicas that have their status attribute set to error, in_sync or out_of_sync. They could even delete an active replica as long as there is another active replica (as could be the case with writable replication style). Before the delete_replica call is made to the driver, an update_access call is made to ensure access rules are safely removed for the replica.

Administrators may also force-delete replicas. Any driver exceptions will only be logged and not re-raised; the replica will be purged from manila’s database.

Promoting a replica¶

Users can promote replicas that have their replica_state attribute set to in_sync. Administrators can attempt to promote replicas that have their replica_state attribute set to out_of_sync or error. During a promotion, if the driver raises an exception, all replicas will have their status attribute set to error and recovery from this state will require administrator intervention.

Resyncing a replica¶

Prior to a planned failover, an administrator could attempt to update the data on the replica. The update_replica_state call will be made during such an action, giving drivers an opportunity to push the latest updates from the active replica to the secondaries.

Creating a snapshot¶

When a user takes a snapshot of a share that has replicas, manila creates as many snapshot instances as there are share replicas. These snapshot instances all begin with their status attribute set to creating. The driver is expected to create the snapshot of the active replica and then begin to replicate this snapshot as soon as the active replica’s snapshot instance is created and becomes available.

Deleting a snapshot¶

When a user deletes a snapshot, the snapshot instances corresponding to each replica of the share have their status attribute set to deleting. Drivers must update their secondaries as soon as the active replica’s snapshot instance is deleted.

Driver Interfaces¶

As part of the _update_share_stats() call, the base driver reports the replication_domain capability. Drivers are expected to update the replication_type capability.

Drivers must implement the methods enumerated below in order to support replication. promote_replica, update_replica_state and update_replicated_snapshot need not be implemented by drivers that support the writable style of replication. The snapshot methods create_replicated_snapshot, delete_replicated_snapshot and update_replicated_snapshot need not be implemented by a driver that does not support snapshots.

Each driver request is made on a specific host. Create/delete operations on secondary replicas are always made on the destination host. Create/delete operations on snapshots are always made on the active replica’s host. update_replica_state and update_replicated_snapshot calls are made on the host that the replica or snapshot resides on.

Replicated Snapshot interfaces:¶

class ShareDriver(driver_handles_share_servers, *args, **kwargs)

Class defines interface of NAS driver.

create_replicated_snapshot(context, replica_list, replica_snapshots, share_server=None)

Create a snapshot on active instance and update across the replicas.

Note

This call is made on the ‘active’ replica’s host. Drivers are expected to transfer the snapshot created to the respective replicas.

The driver is expected to return model updates to the share manager. If it was able to confirm the creation of any number of the snapshot instances passed in this interface, it can set their status to ‘available’ as a cue for the share manager to set the progress attr to ‘100%’.

Parameters:

context – Current context
replica_list – List of all replicas for a particular share The ‘active’ replica will have its ‘replica_state’ attr set to ‘active’.

Example:

[
    {
    'id': 'd487b88d-e428-4230-a465-a800c2cce5f8',
    'share_id': 'f0e4bb5e-65f0-11e5-9d70-feff819cdc9f',
    'replica_state': 'in_sync',
        ...
    'share_server_id': '4ce78e7b-0ef6-4730-ac2a-fd2defefbd05',
    'share_server': <models.ShareServer> or None,
    },
    {
    'id': '10e49c3e-aca9-483b-8c2d-1c337b38d6af',
    'share_id': 'f0e4bb5e-65f0-11e5-9d70-feff819cdc9f',
    'replica_state': 'active',
        ...
    'share_server_id': 'f63629b3-e126-4448-bec2-03f788f76094',
    'share_server': <models.ShareServer> or None,
    },
    ...
]

Parameters:: replica_snapshots – List of dictionaries of snapshot instances. These snapshot instances track the snapshot across the replicas. All the instances will have their status attribute set to ‘creating’.

Example:

 [
    {
    'id': 'd3931a93-3984-421e-a9e7-d9f71895450a',
    'snapshot_id': '13ee5cb5-fc53-4539-9431-d983b56c5c40',
    'status: 'creating',
    'progress': '0%',
        ...
    },
    {
    'id': '8bda791c-7bb6-4e7b-9b64-fefff85ff13e',
    'snapshot_id': '13ee5cb5-fc53-4539-9431-d983b56c5c40',
    'status: 'creating',
    'progress': '0%',
        ...
    },
    ...
]

Parameters:: share_server – <models.ShareServer> or None
Returns:: List of dictionaries of snapshot instances. The dictionaries can contain values that need to be updated on the database for the snapshot instances being created.
Raises:: Exception. Any exception in this method will set all instances to ‘error’.

delete_replicated_snapshot(context, replica_list, replica_snapshots, share_server=None)

Delete a snapshot by deleting its instances across the replicas.

Note

This call is made on the ‘active’ replica’s host, since drivers may not be able to delete the snapshot from an individual replica.

The driver is expected to return model updates to the share manager. If it was able to confirm the removal of any number of the snapshot instances passed in this interface, it can set their status to ‘deleted’ as a cue for the share manager to clean up that instance from the database.

Parameters:

context – Current context
replica_list – List of all replicas for a particular share The ‘active’ replica will have its ‘replica_state’ attr set to ‘active’.

Example:

[
    {
    'id': 'd487b88d-e428-4230-a465-a800c2cce5f8',
    'share_id': 'f0e4bb5e-65f0-11e5-9d70-feff819cdc9f',
    'replica_state': 'in_sync',
        ...
    'share_server_id': '4ce78e7b-0ef6-4730-ac2a-fd2defefbd05',
    'share_server': <models.ShareServer> or None,
    },
    {
    'id': '10e49c3e-aca9-483b-8c2d-1c337b38d6af',
    'share_id': 'f0e4bb5e-65f0-11e5-9d70-feff819cdc9f',
    'replica_state': 'active',
        ...
    'share_server_id': 'f63629b3-e126-4448-bec2-03f788f76094',
    'share_server': <models.ShareServer> or None,
    },
    ...
]

Parameters:: replica_snapshots – List of dictionaries of snapshot instances. These snapshot instances track the snapshot across the replicas. All the instances will have their status attribute set to ‘deleting’.

Example:

 [
    {
    'id': 'd3931a93-3984-421e-a9e7-d9f71895450a',
    'snapshot_id': '13ee5cb5-fc53-4539-9431-d983b56c5c40',
    'status': 'deleting',
    'progress': '100%',
        ...
    },
    {
    'id': '8bda791c-7bb6-4e7b-9b64-fefff85ff13e',
    'snapshot_id': '13ee5cb5-fc53-4539-9431-d983b56c5c40',
    'status: 'deleting',
    'progress': '100%',
        ...
    },
    ...
]

Parameters:: share_server – <models.ShareServer> or None
Returns:: List of dictionaries of snapshot instances. The dictionaries can contain values that need to be updated on the database for the snapshot instances being deleted. To confirm the deletion of the snapshot instance, set the ‘status’ attribute of the instance to ‘deleted’ (constants.STATUS_DELETED)
Raises:: Exception. Any exception in this method will set the status attribute of all snapshot instances to ‘error_deleting’.

update_replicated_snapshot(context, replica_list, share_replica, replica_snapshots, replica_snapshot, share_server=None)

Update the status of a snapshot instance that lives on a replica.

Note

For DR and Readable styles of replication, this call is made on the replica’s host and not the ‘active’ replica’s host.

This method is called periodically by the share manager. It will query for snapshot instances that track the parent snapshot across non-‘active’ replicas. Drivers can expect the status of the instance to be ‘creating’ or ‘deleting’. If the driver sees that a snapshot instance has been removed from the replica’s backend and the instance status was set to ‘deleting’, it is expected to raise a SnapshotResourceNotFound exception. All other exceptions will set the snapshot instance status to ‘error’. If the instance was not in ‘deleting’ state, raising a SnapshotResourceNotFound will set the instance status to ‘error’.

Parameters:

context – Current context
replica_list – List of all replicas for a particular share The ‘active’ replica will have its ‘replica_state’ attr set to ‘active’.

Example:

[
     {
      'id': 'd487b88d-e428-4230-a465-a800c2cce5f8',
      'share_id': 'f0e4bb5e-65f0-11e5-9d70-feff819cdc9f',
      'replica_state': 'in_sync',
      ...
      'share_server_id': '4ce78e7b-0ef6-4730-ac2a-fd2defefbd05',
      'share_server': <models.ShareServer> or None,
     },
     {
      'id': '10e49c3e-aca9-483b-8c2d-1c337b38d6af',
      'share_id': 'f0e4bb5e-65f0-11e5-9d70-feff819cdc9f',
      'replica_state': 'active',
      ...
      'share_server_id': 'f63629b3-e126-4448-bec2-03f788f76094',
      'share_server': <models.ShareServer> or None,
     },
      ...
]

Parameters:: share_replica – Share replica dictionary. This replica is associated with the snapshot instance whose status is being updated. Replicas in ‘active’ replica_state will not be passed via this parameter.

Example:

{
    'id': 'd487b88d-e428-4230-a465-a800c2cce5f8',
    'share_id': 'f0e4bb5e-65f0-11e5-9d70-feff819cdc9f',
    'deleted': False,
    'host': 'openstack2@cmodeSSVMNFS1',
    'status': 'available',
    'scheduled_at': datetime.datetime(2015, 8, 10, 0, 5, 58),
    'launched_at': datetime.datetime(2015, 8, 10, 0, 5, 58),
    'terminated_at': None,
    'replica_state': 'in_sync',
    'availability_zone_id': 'e2c2db5c-cb2f-4697-9966-c06fb200cb80',
    'export_locations': [
        models.ShareInstanceExportLocations,
    ],
    'access_rules_status': 'in_sync',
    'share_network_id': '4ccd5318-65f1-11e5-9d70-feff819cdc9f',
    'share_server_id': '4ce78e7b-0ef6-4730-ac2a-fd2defefbd05',
}

Parameters:: replica_snapshots – List of dictionaries of snapshot instances. These snapshot instances track the snapshot across the replicas. This will include the snapshot instance being updated as well.

Example:

 [
    {
    'id': 'd3931a93-3984-421e-a9e7-d9f71895450a',
    'snapshot_id': '13ee5cb5-fc53-4539-9431-d983b56c5c40',
        ...
    },
    {
    'id': '8bda791c-7bb6-4e7b-9b64-fefff85ff13e',
    'snapshot_id': '13ee5cb5-fc53-4539-9431-d983b56c5c40',
        ...
    },
    ...
]

Parameters:: replica_snapshot – Dictionary of the snapshot instance. This is the instance to be updated. It will be in ‘creating’ or ‘deleting’ state when sent via this parameter.

Example:

{
    'name': 'share-snapshot-18825630-574f-4912-93bb-af4611ef35a2',
    'share_id': 'd487b88d-e428-4230-a465-a800c2cce5f8',
    'share_name': 'share-d487b88d-e428-4230-a465-a800c2cce5f8',
    'status': 'creating',
    'id': '18825630-574f-4912-93bb-af4611ef35a2',
    'deleted': False,
    'created_at': datetime.datetime(2016, 8, 3, 0, 5, 58),
    'share': <models.ShareInstance>,
    'updated_at': datetime.datetime(2016, 8, 3, 0, 5, 58),
    'share_instance_id': 'd487b88d-e428-4230-a465-a800c2cce5f8',
    'snapshot_id': '13ee5cb5-fc53-4539-9431-d983b56c5c40',
    'progress': '0%',
    'deleted_at': None,
    'provider_location': None,
}

Parameters:: share_server – <models.ShareServer> or None
Returns:: replica_snapshot_model_update: a dictionary. The dictionary must contain values that need to be updated on the database for the snapshot instance that represents the snapshot on the replica.
Raises:: exception.SnapshotResourceNotFound Raise this exception for snapshots that are not found on the backend and their status was ‘deleting’.

Share Replication

manila 18.1.0.dev41

Page Contents

Share Replication

Share Replication¶

Storage availability zones and replication domains¶

Replication types¶

Health of a share replica¶

Promotion¶

Snapshots¶

Access Rules¶

Understanding Replication Workflows¶

Creating a share that supports replication¶

Creating a replica¶

Deleting a replica¶

Promoting a replica¶

Resyncing a replica¶

Creating a snapshot¶

Deleting a snapshot¶

Driver Interfaces¶

Share Replica interfaces:¶

Replicated Snapshot interfaces:¶

manila 18.1.0.dev41

Page Contents