Replicating a consistency group
The consistency group is replicated as a whole. The replica is comprised of a source consistency group and a target consistency group. All of the consistency group's members are replicated, and the replica state of the consistency group applies to all of its members.
The replica RPO
For instructions on how to set the RPO, see: How to set the replica RPO and interval.
The RPO (Recovery Point Objective) is the maximum time the replica is allowed to lag. The RPO is configured with another parameter - the Sync Interval. The sync job is expected to be completed prior to the completion of the Sync Interval. The RPO must be greater than the Sync Interval. The best practice is to set the RPO to be twice the time of the Sync Interval.
The RPO is measured in seconds.
- RPO OK – The lag between the source and target < RPO
- RPO LAGGING - The lag between the source and target > RPO
- N/A – RPO not set
The system creates events for changes to the RPO status.
Adding a member to a replicated consistency group
As a consistency group always has all of its members replicated (or all of them not replicated), you can add a member to a consistency group using the following guidelines:
- You can add a non-replicated dataset to a non-replicated consistency group.
- You can add a replicated dataset to a replicated consistency group.
- You cannot add a replicated dataset to a non-replicated consistency group.
To add a dataset to a replicated consistency group, add the source dataset to the source consistency group. InfiniBox will automatically replicate it to the target consistency group.
The consistency group settings (Sync Interval and RPO) are imposed on the added member.
Impact on the consistency group state
The added member undergoes initialization upon the next sync job. This causes the entire consistency group to undergo an initialization. A replicated consistency group is usually in either replicating or idle state. Adding an volume that is not yet replicated to the consistency group forces the entire consistency group to yield to an initializing state. This state is imposed on the consistency group even if all of its members (except for the new member) are synchronized with the target.
A consistency group that is in initialization state cannot have its role changed in case of a disaster recovery. For the role of the consistency group to be changed, the initializing member must be removed - losing any advancement in replicating to the target member - so that the consistency group can be moved out of the initializing state and the role can be changed.
To prevent this situation, consider replicating the new member separately, prior to adding it to the consistency group. Only after it is synchronized with the target, add it to the consistency group. When a replicated volume is added to a replicated consistency group, its target is added to the target consistency group.
Removing a member from a replicated consistency group
Removing a member from the consistency group removes the source member from the source consistency group, removes the target member from the target consistency group, and - optionally - retains the replica between the source and target member. If you instruct the replica to be retained, the replica of the removed dataset keeps the settings (Sync Interval and RPO) and the state (synchronized, not-synchronized) of the consistency group. You can remove a member from a consistency group regardless of the state of the consistency group.
Capacity and throughput reporting
- The system reports the capacity consumption of the replication snapshots.
- The system reports the throughput of replica that is currently replicating.