Workload Optimization offers you the potential to automate fully a significant portion of your cluster workload rebalancing tasks. The tasks to accomplish workload automation are as follows:

  1. Configure the Workload Automation Details. See Workload Automation Details.

  2. Tag VMs for cluster placement. See Business Intent - Host-Based Virtual Machine Placement and Business Intent: Tag-Based VM Placement in Clusters.

  3. If you do not use the AUTOMATE function in the Optimization Recommendation pane at the Workload Automation screen, configure the two Workload Optimization alerts to be triggered when cluster CPU/memory limits are breached, and configure them as automated. When the alerts are automated, the actions calculated by Workload Optimization are run automatically. See Configuring Workload Optimization Alerts

Prerequisites

Workload Optimization acts on objects associated with the VMware vSphere Solution that connects vRealize Operations Manager to one or more vCenter Server instances. The virtual objects in this environment include a vCenter Server, data centers and custom data centers, cluster compute and storage resources, host systems, and virtual machines. Specific requirements:

  • A vCenter Adapter configured with the actions enabled for each vCenter Server instance.

  • A vCenter Server instance with at least two datastore clusters with sDRS enabled and fully automated.

  • Any non-datastore clusters must have DRS enabled and fully automated

  • Storage vMotion must be set to ON at Workload Automation Details. The default is On.

  • You must have permission to access all objects in the environment.

Design Considerations

The following rules constrain the possible computer and storage resource moves that can be performed.

Note:

When vRealize Operations Manager suggests that you optimize clusters in a data center, the system does not guarantee it can run an optimization action. vRealize Operations Manager analytics can determine that optimization is desirable and can create a rebalancing plan. However, the system cannot automatically identify all the architectural constraints that may be present. Such constraints may prevent an optimization action, or cause an action in progress to fail.

  • Moving compute and storage resources is allowed only within, not across data centers or custom data centers.

  • Storage resources cannot be moved across non-datastore clusters. Storage can move only across datastore clusters that have sDRS fully automated.

  • Compute-resource-only moves are permitted through shared storage.

  • Virtual machines defined with affinity rules or anti-affinity rules are not to be moved.

  • Virtual machines cannot be moved when residing on a local datastore, unless a storage swap exists on the local datastore.

  • Virtual machines cannot be moved if they have data residing across multiple datastore clusters. Compute-only moves with similar shared storage are not permitted.

  • A virtual machine cannot have data that resides across different storage types. For example, if a virtual machine has a VM disk on a datastore and a second VM disk on a datastore cluster, the virtual machine does not move, even when the datastore is shared with the destination or has swap on it.

  • A virtual machine can use RDM so long as the destination datastore cluster can access the RDM LUN.

  • A virtual machine can implement VM disks on multiple datastores inside a single datastore cluster.

  • Virtual machines cannot be placed inside clusters managed by vRealize Automation.

  • Workload Optimization may suggest moving virtual machines that are protected by vSphere Replication or Array Based Replication. You must ensure that all the clusters within a selected data center or custom data center have replication available. You can set up DRS affinity rules on virtual machines that you do not want moving across clusters.