- Kubeadm: during the preflight check "CreateJob" of "kubeadm upgrade", check if there are no nodes where a Pod can schedule. If there are none, show a warning and skip this preflight check. This can happen in single node clusters where the only node was drained. ([#124503](https://github.com/kubernetes/kubernetes/pull/124503), [@neolit123](https://github.com/neolit123)) [SIG Cluster Lifecycle]

	// Address represents the addresses the service can be reached at.
	// There may be multiple addresses for a single service if it resides in multiple networks,
	// multiple clusters, and/or if it's dual stack (TODO: support dual stack).
	// For a headless kubernetes service, this list will be empty.
	Addresses []*NetworkAddress `protobuf:"bytes,4,rep,name=addresses,proto3" json:"addresses,omitempty"`

  // TODO(patn): move this pass after cluster formation to remove the
  // complexity with replication info and metadata, cluster checking and
  // generalizing to multiple TPU clusters.
  Region* region = (*forward_pass_ops.begin())->getParentRegion();
  StringAttr replication_attr = GetReplicationAttr(*forward_pass_ops.begin());
  llvm::SmallVector<Operation*> checkset(forward_pass_ops.getArrayRef());

	// PodSchedulingContext.*.UnsuitableNodes is complete) but may happen
	// when it doesn't (= PodSchedulingContext.*.UnsuitableNodes had to be
	// truncated).
	//
	// Truncation only happens for very large clusters and then may slow
	// down scheduling, but should not break it completely. This is
	// acceptable while DRA is alpha and will be investigated further
	// before moving DRA to beta.
	if podScheduling.Spec.SelectedNode != "" {