In releases of the VSA prior to 5.x, running vCenter Server in a virtual machine within a VSA cluster is an unsupported configuration.
Starting with VSA 5.x, you may run vCenter on a virtual machine within a 3-node VSA cluster, and within a 2-node VSA cluster under certain conditions. See
Acceptable supported configurations for vSphere Storage Appliance (VSA) clusters (2043783) for more information.
For more information, see How a VSA Cluster Handles Failures in the
VMware vSphere Storage Appliance Installation and Configuration Guide. In addition to this information, here are additional details regarding possible failure and recovery scenarios.
Note: In order for a VSA cluster to be “online” and in a fully-functioning state, three nodes must be online and in communication with one another from a networking perspective.
2-Node Clusters:
- In a 2-node cluster, the VSA Cluster Service is considered the 3rd node.
- If any single node of the three fails, or becomes isolated from the others from a networking point of view, the cluster becomes "vulnerable" to a second failure.
- If the cluster is in such a "vulnerable" state, and the reason the cluster is vulnerable is because the failed / isolated node is the VSA Cluster Service, then the NFS Storage exported by the cluster will remain accessible, and not “degraded” (the storage will still be replicating between the two main nodes).
- If the cluster is in such a “vulnerable” state, and the reason is because the failed / isolated node is one of the VSA Appliances, then the NFS Storage exported by the cluster will remain accessible, but will be in a “degraded” state (the storage will not be replicating between the two nodes, because of the failure of one of the nodes).
- If an additional node fails or becomes isolated when the cluster is in a “vulnerable” state, the cluster will go “offline”. This means the NFS datastores will neither be accessible, nor will they be replicated. This condition will continue until one of the two failed or isolated nodes returns to communication with the surviving node, at which point the cluster will return to “online” status. If the first node back is the VSA Cluster Service, then the NFS datastores will become accessible, but will remain in a “degraded” status. The NFS datastores will stay in a “degraded” status until full synchronization tasks are completed.
3-Node Clusters:
- If any single node of the three fails, or becomes isolated from the others from a networking point of view, the cluster becomes “vulnerable” to a second failure.
- If the cluster is in such a “vulnerable” state, then the NFS Storage exported by the cluster will remain accessible, but two of the three datastores will become “degraded” (the storage replication will be impaired, because of the failure of one of the nodes).
- If an additional node fails or becomes isolated when the cluster is in a “vulnerable” state, the cluster will go “offline”. This means the NFS datastores will neither be accessible, nor will they be replicated. This condition will continue until one of the two failed or isolated nodes returns to communication with the surviving node, at which point the cluster will return to “online” status. The NFS datastores will stay in a “degraded” status until full synchronization tasks are completed.
Please keep in mind that if you are running vCenter in a virtual machine on NFS Storage exported by the cluster, vCenter will be offline if the cluster is offline.
Also, the VSA Manager GUI is integrated with vCenter via a plug-in, and so certain operations are not possible or straightforward when the vCenter Server is running within VSA-provisioned storage.
The most problematic operation is Cluster Maintenance Mode, which cannot be entered in such a configuration because the operation requires all VMs resident on the VSA to be powered down. Since this would include the vCenter Server, VSA Manager would be inaccessible as a consequence of the power-down.
VSA manageability and troubleshooting can be more difficult if vCenter server is stored on VSA-provisioned storage. However, there are benefits to having vCenter inside the VSA Cluster, in terms of HA and vMotion.
In order to retain the option of performing Cluster Maintenance Mode, we suggest that sufficient disk space be reserved on the hosts’ local storage, to be able to move vCenter out of the cluster storage before such an operation is started.