New ESXi Hosts Zoned to One PowerStore Node
An eight-host VMware estate is connected to a PowerStore 3200T through two FC fabrics. Each host has two HBAs, and each node has four 16 Gb FC ports. The storage team adds two new ESXi hosts and creates eight 2 TB volumes for them. The new hosts see four paths, all logged in to node A; the SAN admin copied an old zone set that contained only node A target ports. Within a day, the new hosts report 8–12 ms latency while the existing hosts stay near 1.5 ms. PowerStore Manager shows node A front-end CPU at 88%, node B at 32%, and inter-node link utilization above 70%. The volumes are healthy, and no alerts are raised. The application owner wants the new hosts moved to node A volumes to 'keep traffic local'.
1. What is the most likely cause, and why does it affect only part of the new hosts' I/O?
Read the worked answer
The new hosts are zoned only to node A target ports. PowerStore is active-active at the appliance level, but each volume has a node that owns it and provides the optimal path. With all logged-in paths landing on node A, I/O to volumes owned by node B must cross the inter-node link. That explains why only part of the workload is slow, why node A CPU is high while node B is relatively idle, and why inter-node utilization is elevated. The four paths are not proof of redundancy; they are four paths to the same node. Existing hosts are zoned to both nodes and therefore use optimal paths for all volumes.
2. What changes would you make to zoning and host configuration, and how would you verify success?
Read the worked answer
Add zones on both fabrics from each new host HBA to the node B FC ports, keeping the existing node A zones. Do not replace one set with another. Rescan the HBAs and confirm that each host sees paths to both nodes; check esxcli storage nmp device list and path list to confirm all paths are active and the PSP is using them. In PowerStore Manager, confirm the host initiators are registered and the volumes are mapped to the host. Then check node CPU balance and host latency; expect node A CPU to fall and inter-node traffic to drop. Also verify that no single point of failure remains in the zone set.
3. Should you instead migrate the new volumes to node A? Justify your answer.
Read the worked answer
No. Moving the volumes to node A would make the new hosts optimal but would concentrate load on node A and leave node B underused. It also reduces resiliency because a node failure would affect more volumes. The correct fix is path locality: zone to both nodes and let PowerStore place volumes normally. If a temporary migration is used, it should be for a specific capacity or performance reason, not to compensate for a one-node zone design.