About the client and policy assignments with disk volumes in MVG volume
MVG server uses client and policy combinations to distribute backup data across disk volumes. One combination can be assigned to one or multiple disk volumes with a priority order. The default disk volume is called the primary node, and others are called historic nodes.
When a client and policy combination is used for the first time, MVG selects a disk volume based on certain criteria. See See MVG load balancing levels. The subsequent jobs continue with the same client and policy combination using the selected disk volume unless it becomes unavailable or full.
If the primary node becomes unavailable, MVG does not immediately switch to another volume during the next backup job. Instead, it waits for a predefined duration to determine whether the outage is temporary. This delay may result in job failures before the threshold is reached, but it helps avoid unnecessary data duplication and preserves deduplication efficiency. Once MVG decides to use a different volume, it selects the next available historic node if one exists. If not, a new disk volume is assigned as a historic node for that client/policy combination. When the original primary node becomes available again, MVG resumes using it for future jobs.
See About the node failure management.
If the primary node becomes full, MVG immediately switches to the next available historic node or assigns a new disk volume as the new primary node. The previously full volume becomes a historic node. This is the only case where the primary node is changed automatically.
See Specifications and default values of the MVG tuning parameters.
For VLDB and MSDPLB+ policies with multistreaming enabled, the MVG server selects multiple disk volumes for the streams of the client and policy combinations. The streams are distributed with round robin to the list of the primary and historic nodes. It is useful when the data source is large, or a better performance is required. However, it does not contribute to the better deduplication rate.
It is recommended to configure the number of streams as a multiple of the number of disk volumes (for example, 4, 8, 12 for 4 volumes) to ensure even distribution. Any change in stream count or disk volume count affects deduplication in the first backup job after the change.
Assignments can be managed manually to control data distribution and failover behavior. You can predefine primary and historic nodes before a client/policy combination is used.
To manage the assignments of the client and policy combinations with the disk volumes:
- Get DSID of the MVG volume.
/usr/openv/pdde/pdcr/bin/cacontrol - mvg listvvols
- Get assigned nodes for a client/policy combination.
/usr/openv/pdde/pdcr/bin/cacontrol --cluster get-cp-assignment <dsid> [<client> [<policy>]]
Output format:
<client-name>/<policy-name>: <primary-node>[,<historic-nodes…>]
Example:
mvgdemo-host1.mvgdemo.com/host4-local-pl01: mvgdemo-host4.mvgdemo.com,mvgdemo-host2.mvgdemo.com
- Change the primary node of a client/policy combination.
/usr/openv/pdde/pdcr/bin/cacontrol --cluster set-cp-assignment <dsid> <client> <policy> <msdp-server>
Example:
cacontrol --cluster get-cp-assignment 3 mvgdemo-host1.mvgdemo.com mvg-local-pl11 mvgdemo-host1.mvgdemo.com/mvg-local-pl11: mvgdemo-host2.mvgdemo.com,mvgdemo-host3.mvgdemo.com Received 1 client-policy assignments
Use the above command to change the primary node that is assigned to a client/policy combination.
To add a historic node, run the command twice. First, assign the new node as the primary. Then, reassign the original node as the primary. This action moves the newly added node to the historic node list.
You can also use the command to predefine both primary and historic nodes before the client and policy combination is used. This approach helps you control data distribution and failover behavior more predictably.