This section describes how to configure replication on the Core. As of version 4.3, you can configure physical or virtual Cores for seamless failover and recovery between data centers without any data loss. In this environment, your primary and secondary data centers are always synchronized in case of large scale failures such as power loss, natural disasters, or hardware failure. You connect two separate Cores that are each connected to their own storage array in two separate data centers.

The primary data center receives all the reads and writes from the Edge, and is connected to the secondary data center. The secondary data center hosts the replica LUNs, which are copies of the primary LUNs located at the primary data center.

Replication can interoperate with all existing physical or virtual Cores, and all existing Edges, as long as the software version is compatible (version 4.0 and later).

This section describes the following topics:

You must meet the following requirements to set up replication:

This table describes the basic steps needed to set up replication across data centers, followed by detailed procedures.

First, set up the primary data center, where the Edges and LUNs will reside. The secondary data center is on standby and should not have any configured LUNs or Edges, and it is the data center that the primary will be replicating to.

Next, set up the secondary data center.

Once you have set up the data centers, they are prepared to connect to each other. If the Cores are set up for high-availability, repeat these steps on the failover peer.

The next step is to peer the Cores together from the primary Core.

After you have specified the roles of the primary and secondary Cores for replication, pair them together from the primary data center to establish connections for replication. If you have two Cores in a high-availability configuration, you pair both of them to the secondary data center. For example, you could pair Core X to Core X’ and Core Y to Core Y’.

After you have paired the Cores, the next step is to set up the Witness.

The Witness is an Edge that you choose to be the authoritative source on each data center’s state in case they enter a “split-brain” scenario, in which both Cores attempt to journal writes at the same time. Any requests to suspend replication and start journaling are approved by the Witness, which ensures that only one data center is approved for journaling at any given time.

You can only configure the Witness from the primary data center; however, once it is set up, it is available and visible on all Cores across both the primary and secondary data centers.

These requirements must be met for the Witness to work:

If the Cores are set up for high-availability, one Core has the role of the leader and the other has the role of the follower. Some configuration changes are only possible on the leader Core. If the leader is down, the follower Core assumes the role of leader. When the original leader comes back up again, it resumes its role.

Even though you can add Edges to both high-availability Cores, only the Edges on the leader Core can be configured as a Witness.

Now that the Cores are aware of each other and the Witness is configured, add the replica LUNs on the primary data center and start first sync.

If the Witness is down or experiencing another type of failure, you can change it to a different Edge in the Witness Information panel on the primary Core. Any issues relating to the Witness appear in the Reports > Diagnostics: Alarm Status page under Edge Service.

After you have set up the Witness, you will map the replica LUNs. A replica LUN is a LUN on the secondary data center that acts as a mirror to the primary LUN in the primary data center. The replica LUN can be either iSCSI or Fibre Channel, and must be equal in size or larger than the primary LUN (a 1 GB leeway exists). To start replication for an Edge, each of its mapped LUNs must have an associated replica LUN. When you start replication for an Edge, the Core replicates all the LUNs that are mapped to that Edge.

Once the replication for an Edge is enabled, the first sync of data from the primary LUN to the replica LUN starts. During this process, the Core continues to run normally without interruption. There are two options for first sync:

After the LUN is copied to the secondary data center during the first sync, the primary LUNs and replica LUNs are in a consistent state and replication is active. You can choose a different sync type for each LUN.

During an extended data center failure or planned network event such as a network upgrade on the secondary data center, you can suspend replication on the primary data center to prevent the Edge’s blockstore from accumulating excessive data. Suspending replication will not clear any replication settings.

While replication is suspended, the primary data center Core uses the Journal LUN to track the writes from the Edge until the connection is restored. When the secondary data center is restored, the Cores automatically reconnect and journaled writes are synced to the secondary data center to make the replica LUN and primary LUN consistent. Failback is possible if this data center is preferred. For more information about failback, see Failing Back to the Primary Data Center.

If the Cores are set up for high-availability, replication must be suspended from the leader Core in the primary data center. Once the leader approves the change, the follower also suspends replication, and they are both in the suspended state. If the leader fails for any reason, the follower automatically becomes the leader.

In the event of primary data center failure, where the data center is decommissioned for an extended period of time, you can easily have the primary data center’s LUNs and Edges fail over to the secondary data center to continue Edge operations. For the Edges, this process is entirely transparent. A failover is not automatic; it is initiated by the administrator and is only permitted when the following criteria are met:

After failover is complete, the Secondary Core’s role in Replication Configuration panel will change to Primary. All Edge connections and data commits will move to the new Primary, Edge commits will be “In Progress”, and replication will be suspended.

The original secondary data center will now act as primary and the original primary data center will act as secondary whenever it recovers from the failure.

Once a failed data center has recovered, you can restore its role to primary if it is a preferred data center (due to hardware configuration or geographical location, for example). To failback to the original primary data center, you initiate failover from the preferred data center, which is currently in the secondary role.

To stop replication for an Edge, you must first terminate replication for the Edge, and then unmap all replica LUNs associated with that Edge. To completely terminate replication for a Core, you must disable replication for all the replicating Edges.