Skip to main content

Drift Detection

Purpose: For operators, explains what opencenter cluster drift compares today and which providers actually participate in the infrastructure drift contract.

What the Command Compares

The drift workflow builds a desired infrastructure model from the cluster configuration and compares it with the provider's live API state.

Today that means:

  • OpenStack: servers, networks, security groups, load balancers, volumes, and floating IPs
  • VMware: configured VM nodes, attached networks, and datastore-backed storage expectations

Supported Providers

ProviderStatusReconciliation
OpenStackSupportedLimited safe reconciliation for mutable infrastructure items
VMwareSupportedDetection only; remediation is manual
KindNot applicableNo infrastructure drift backend
BaremetalNot applicableNo infrastructure drift backend
AWSRemoved from GA drift registryNot supported

Typical Flow

# Detect drift
opencenter cluster drift detect prod-cluster

# Filter output by severity
opencenter cluster drift detect prod-cluster --severity=critical

# Preview any supported reconciliation
opencenter cluster drift reconcile prod-cluster --dry-run

VMware-Specific Behavior

VMware drift detection uses the configured vCenter metadata plus secrets.vsphere_csi credentials to inspect:

  • the configured datacenter
  • the named VM nodes
  • the networks attached to those VMs
  • the datastores backing those VMs

The command reports differences, but it does not mutate vSphere resources for you.

Why Kind and Baremetal Are Different

Kind is a local lifecycle provider, not a cloud-resource backend. Baremetal relies on pre-provisioned hosts that openCenter does not create or own. In both cases, an infrastructure drift contract would be misleading, so those providers are intentionally excluded.