vSAN Stretched Cluster recovery fails with ESA and OSA architecture mismatch
Article ID: 444762
Updated On:
Products
Issue/Introduction
Symptoms After accidentally deleting and attempting to recreate a vSAN stretched cluster, VMs remain inaccessible and the following errors appear in Skyline Health:
vSAN ESA is disabled on the cluster but is enabled on this host
vSAN is enabled on the cluster without vSAN ESA, but a storage pool is detected on this host
Persistent network isolation warnings or MTU profile mismatches.
Environment
VMware vSAN 8.0 and higher
Stretched Cluster Configuration
Mixed Storage Architectures (ESA and OSA)
Cause
The issue occurs when a vSAN Express Storage Architecture (ESA) Witness Appliance is deployed to a cluster consisting of Original Storage Architecture (OSA) data nodes. vSAN requires architectural consistency across all nodes, including the witness. An ESA witness automatically creates storage pools that are explicitly incompatible with OSA disk group topologies.
Resolution
To restore the cluster to a healthy state, align the storage architecture of the witness host with the data nodes:
- Evict the current ESA witness appliance from the vSAN stretched cluster configuration.
- Deploy a new vSAN Witness Appliance that specifically matches the architecture of the data nodes (OSA).
- Ensure the ESXi build version of the witness appliance matches the build version of the data nodes.
- Configure the MTU on the witness host VMkernel ports to match the data nodes (e.g., MTU ####).
- Run the Stretched Cluster configuration wizard to bind the new OSA witness to the cluster.
- Perform a disk format upgrade if prompted once the cluster health is "Green."
If the issue persists, contact Broadcom Support by following the steps in Contact Support.
Additional Information
Feedback
