WSFC node reboot triggers pRDM SCSI resets

search cancel

book

calendar_today

VMware vSphere ESXi VMware vSphere ESX 7.x

Windows Server Failover Clustering (WSFC) is configured using physical RDMs (pRDM) and Physical Bus sharing, while both nodes reside within a single ESXi host (Cluster in a Box/CIB).
A running cluster node experiences I/O failures or storage connectivity drops when the partner node initiates a reset or reboot.
During reboot of one of the WSFC nodes, the /var/log/vmkernel.log file reports the following events:
- "Attempt to issue lun reset on device..."
- Against shared RDM LUNs: "H:0x0 D:0x2 P:0x0 Valid sense data: 0x6 0x29 0x0" (Power on reset)

Placing all nodes of a WSFC on the same physical ESXi host, known as a "Cluster-in-a-Box" (CiB) configuration, is unsupported in production when using shared physical RDM disks.
When nodes share the same physical host, they also share the same physical storage queues and HBA paths.
A reboot or SCSI reset initiated by one node can disrupt the storage traffic for the partner node, leading to I/O interruptions and cluster instability.

To ensure a supported environment, all nodes of a WSFC using shared disks must be deployed in a "Cluster Across Boxes" (CAB) configuration.

Verify that all virtual machines (VMs) acting as cluster nodes are placed on different physical ESXi hosts.
Enforce this placement using VMware Distributed Resource Scheduler (DRS) "Must" Anti-Affinity rules.
- Navigate to vCenter Server > Configure > VM/Host Groups.
- Create a new VM Anti-Affinity Rule.
- Select the virtual machines that serve as cluster nodes.
- Ensure the rule is set to Separate Virtual Machines.

thumb_up Yes

thumb_down No