PowerStore: Metro Volume Polarized During Node Reboot if Data Connection Not Fully Redundant

The data connection between the two appliances participating in Metro Volume may be intentionally configured with a partial data connection, for example, due to environmental limitations.

The data connection state for the Remote System is displayed as Partial_Data_Connection_Loss.

The following diagram illustrates two configurations:

• One with correct and fully redundant data connections
• One with partial data connectivity

If the data connection is partially connected and a node is rebooted while replication I/O is in-flight, the corresponding Metro Volume may become polarized and autorecovers.

An alert is raised for the Metro Volume polarization and fracture. Host I/O is only available on the Metro Volume that wins polarization.

If the host setup is nonunified, workloads running on the losing side of the mirror must restart on the winning side.

For more information about failure scenarios, see the failure scenario table for Metro Volumes (with or without witness) in the Metro Volume whitepaper Info hub.

When I/O is directed to a Metro Volume, the affinity node handles replication I/O.
For example, Node A on the Preferred (P) side and Node B on the Non-Preferred (NP) side may both have Node Affinity for a Metro Volume.

When a node is rebooted while host I/O is in-flight, Metro Volumes may fracture depending on the data connection state.
If a reboot of Node B on the NP side:

• If there is a fully redundant data connection:
  Node A on the P site can still use the data connection between P site Node A and NP site Node B to complete the I/O.
  → The Metro Volume remains in-sync and active.

• If there is only a partial data connection between the two appliances (as mentioned above):
  Node A on the P site has no available data connection to the NP site.
  → Node A cannot replicate the in-flight I/O to the NP side, and the corresponding Metro Volume becomes polarized.

If there is no in-flight I/O to a Metro Volume, the Metro Volume may not be affected.

PowerStore automatically attempts to restore the Metro Volume after the node boots up and the data connection is reestablished.

This is expected behavior.

Dell highly recommends configuring replication with redundant network connections, ensuring that each node can communicate with all nodes on the remote site.

If a Metro Volume fractures during a Non-Disruptive Upgrade (NDU) automatic recovery may be delayed until the update completes.