Home > Storage > PowerStore > Data Protection > Dell PowerStore: Metro Volume > Failure scenarios
This section covers possible failure scenarios with non-uniform and uniform host connectivity. Due to the polarization mechanism, a Metro Volume is only active-active across both PowerStore clusters when the PowerStore clusters are operational, and the links for replication management traffic and replication data traffic are established. During a link failure required for replication, the non-preferred side loses communication with the preferred side of the Metro Volume and switches to offline for the hosts while preferred volumes can still be served from the same PowerStore cluster. During an array failure, only preferred volumes remain accessible without a manual promote of non-preferred volumes on the surviving array. The following subsections show failure scenarios in non-uniform and uniform host configurations.
In this scenario, VMs on host 2 access two different Metro Volumes on PowerStore-B. VM1 is on Metro Volume 1 with the non-preferred role, and VM 2 is on Metro Volume 2 with the preferred role (see below).
In the following figure, due to a link failure, Host 2 loses access to Metro Volume 1 with an all paths down (APD) event, and VM 1 is interrupted. VM 2 continues to run on the preferred Metro Volume 2 mapped to host 2. Host 1 has access to the preferred volume of Metro Volume 1, vSphere HA recognizes that the Volume is still available, and it restarts VM 1 on host 1.
With the same configuration when PowerStore-B is down during a failure situation (array failure), Metro Volume 2 is not accessible until a manual promote operation completes on PowerStore-A. After the promote action of Metro Volume 2 on host 1, the volume and active paths are enabled, and VM 2 could start on host 1.
The example configuration for a uniform configuration is identical to the previous example with extra paths from the hosts to the remote PowerStore cluster. The redundant active paths to local and remote PowerStore clusters give an extra level of high availability. Depending on the host configuration as described in the Host configuration section, the paths to the volume served by remote PowerStore can be ALUA active-non-optimized only, or also active-optimized with active-non-optimized paths dependent on volume node affinity setting. In example as shown below, VM 1, and VM 2 are running on host 2 using different Metro Volumes. While VM 1 uses Metro Volume 1 in the non-preferred role which goes offline during a failure, VM 2 uses Metro Volume 2 in the preferred role.
In a failure scenario, when the link between PowerStore-A and PowerStore-B is affected, host 2 would lose the volume for VM 1. As active paths are still available to the same volume on PowerStore-A, VM 1 remains online without interruption. VMware Native Multipathing (NMP) handles the failover to an available active path.
During an array failure of PowerStore-B, VM 2 would also lose the active path to Metro Volume 2, but it cannot continue on PowerStore-A because Metro Volume 2 is in the non-preferred role. VM 2 can only start with Metro Volume 2 on PowerStore A after a manual Metro Volume promote action is performed on PowerStore-A.