Future saturation recommendation

The Future saturation recommendation detects entities for which the resource usage has either already exceeded thresholds or is expected to exceed the thresholds.

The recommendation is generated by evaluating the natural (organic) growth of resource usage in your cluster due to workload growth, or due to deployment of new projects. It is based on historical trending, and extrapolation to the future. It compares the estimated resource utilization with thresholds.


 Resolution pointers for upward trend scenarios in vSphere clusters

Based on what might be causing the resource usage to trend upward, you can take specific steps to resolve the problem. Here are some examples of the upward trend scenarios and pointers to resolve them:


Upward trend scenarioPointers
If any of the hosts in the vSphere cluster are disconnected or are in maintenance modeTroubleshoot the hosts and recover cluster resources.
If the cluster has the High Availability (HA) option turned ON
  • Ensure that none of the hosts is in a Host unreachable state. If they are, it can affect the apparent CPU utilization of the cluster.
  • Ensure that none of the hosts has an HA Agent in one of the following states:
    • Uninitialized state: A host could be in this state because it is unable to access the datastores of the cluster.
    • Initialization Error state
    • Uninitialization Error state
    • Host Failed state
  • Ensure that the HA cluster has not declared any of the hosts as failed. This could happen if the primary (master) host is unable to communicate with a host and is using the datastore heartbeat to verify the connection (because the datastore is not connected properly).

In all these cases, troubleshooting the offending hosts should enable you to recover the resources of the cluster and eliminate the future saturation condition.

  • If the cluster has the Distributed Resource Scheduler (DRS) turned on
  • Additional hosts are available to be powered on, but they are not
  • DRS rules may be preventing VMs from being moved to the available host
  • VMs may be pinned to their current hosts
  • DRS or Distributed Power Management (DPM) might be set to manual mode, which could be causing the power-on recommendations to not be applied.
  • DRS might be disabled on the available host because of a previous failure to exit standby mode.

In all these cases, remove the condition that prevents an available host from powering on to add resources to the cluster and remove the future saturation condition.

If storage is saturating soon or is already saturated
  • Ensure that all datastores in the cluster are visible on the hosts.
  • If there are old snapshots for VMs in the cluster, consider removing them to reclaim disk space.
  • If there are idle VMs or powered off VMs in the cluster, consider removing them to reclaim disk space.
If CPU or memory is saturating soon or is already saturatedIf there are Idle VMs in the cluster, consider removing them to reclaim CPU or memory resources.
If you are familiar with the workload (application) running inside a VMTroubleshoot (for example, stop runaway processes) or tune the application to reduce its footprint, then reduce the reservations or limits available to the VM or its parent resource pools to make the cluster more efficient.
As a last resort, for CPU, memory, or storage
  • Move some VMs to another cluster, especially clusters that are under-subscribed
  • Add another host to the cluster
  • Add CPU, memory, or storage, as recommended or applicable
The recommended actions provide the resource identifiers of the entities for which the resources are saturated or are expected to saturate.


Was this page helpful? Yes No Submitting... Thank you

Comments