vRealize Operations Manager supports continuous availability (CA). CA separates the vRealize Operations Manager cluster into two fault domains, stretching across vSphere clusters, and protects the analytics cluster against the loss of an entire fault domain.

You can configure the analytics cluster with Continuous Availability. This allows the cluster nodes to be stretch across two fault-domains. A fault domain consists of one or more analytics nodes grouped according to their physical location in the data center. With CA, the two fault domains permit vRealize Operations Manager to tolerate failures of an entire physical location and failures from resources dedicated to a single fault domain.

To enable continuous availability within vRealize Operations Manager, the witness node must be deployed in the cluster. The witness node does not collect nor store data. In a situation where network connectivity the two fault-domains is lost, the cluster would go into a split-brain situation. This situation is detected by the Witness Node and one of the fault domains will go offline to avoid data inconsistency issues. You will see a Bring Online button on the admin UI of the nodes which are made offline by the witness node. Before using this option to bring the fault domain online, ensure that the network connectivity between the nodes across the two fault domains is restored and stable. Once confirmed you can bring the fault domain online.

With CA, the data stored in the master node and data nodes grouped in fault domain 1 is always 100% synced to the replica node and data nodes paired in fault domain 2. To enable CA, you must have at least one data node deployed, in addition to the master node. If you have more than one data node, there must be an even number of data nodes including the master node. For example, the cluster must have 2, 4, 6, 8, 10, 12, 14 or 16 nodes based on the appropriate sizing requirements. The data stored in the master node in fault domain 1 is stored and replicated in the replica node in fault domain 2. The data stored in the data nodes in fault domain 1 is stored and replicated in the paired data nodes in fault domain 2. But in case the master node fails, only the replica node can function as the replacement of the master node.

  • CA protects the analytics cluster against the loss of half the analytics nodes specific to one fault domain. You can stretch nodes across vSphere clusters in an attempt to isolate nodes or build failure zones.
  • When CA is enabled, the replica node can take over all functions that the master node provides, in case of a master node failure. The failover to the replica is automatic and requires only two to three minutes of vRealize Operations Manager downtime to resume operations and restart data collection.
    Note: In case of a master node failure, the replica node becomes the master node, and the cluster runs in degraded mode. To fix this, perform any one of the following actions.
    • Correct the master node failure manually.
    • Return to CA mode by replacing the master node. Replacement nodes do not repair the node failure, instead a new node assumes the master node role.
  • In the administration interface, after a CA replica node takes over and becomes the new master node, you cannot remove the previous, offline master node from the cluster. In addition, the previous node remains listed as a master node. To refresh the display and enable the removal of the node, refresh the browser.
  • When CA is enabled, the cluster can survive the loss of half the data nodes, all in one fault domain, without losing any data. CA protects against the loss of only one fault domain at a time. Simultaneously losing data and master/replica nodes, or two or more data nodes in both fault domains, is not supported.
  • A CA enabled cluster will be non-functional if you power off the master node or the master node replica while one of the fault domains is down.
  • When CA is enabled, it lowers the vRealize Operations Manager capacity and processing by half, because CA creates a redundant copy of data throughout the cluster, and the replica backup of the master node. Consider your potential use of CA when planning the number and size of your vRealize Operations Manager cluster nodes. See Sizing the vRealize Operations Manager Cluster.
  • When CA is enabled, deploy analytics cluster nodes, in each fault domain, on separate hosts for redundancy and isolation. You can also use anti-affinity rules that keep nodes on specific hosts in the vSphere clusters.
  • If you cannot keep the nodes separate in each fault domain, you can still enable CA. A host fault might cause the loss of the data nodes in the fault domain, and vRealize Operations Manager can still be available in the other fault domain.
  • If you cannot split the data nodes into different vSphere clusters, do not enable CA. A cluster failure can cause the loss of more than half of the data nodes, which is not supported, and all of vSphere might become unavailable.
  • Without CA, you can keep nodes on the same host in the same vSphere. Without CA, the loss of even one node might make all of vRealize Operations Manager unavailable.
  • When you power off data nodes in both fault domains and change the network settings of the VMs, it affects the IP address of the data nodes. After this point, the CA cluster is no longer accessible and all the nodes status change to "Waiting for analytics". Verify that you have used a static IP address.
  • When you remove a node that has one or more vCenter adapters configured to collect data from a CA-enabled cluster, one or more vCenter adapters associated with that node stops collecting. You must change the adapter configuration to pin them to another node before removing the node.
  • The administration interface displays the resource cache count, which is created for active objects only, but the inventory displays all objects. When you remove a node from a CA-enabled cluster allowing the vCenter adapters to collect data and rebalance each node, the inventory displays a different quantity of objects from that shown in the administration interface.