fix: do not block machine config updates if loadbalancer is down

This check doesn't make sense for the config updates.
We are trying to address the case when the config has broken the API
server and you try to revert it. All api servers are down and you can
not revert it because all api servers are down.

Signed-off-by: Artem Chernyshev <[email protected]>
(cherry picked from commit a0d02ea)

Author: Unix4ever

internal/backend/runtime/omni/controllers/omni/internal/machineset/control_planes_handler.go

@@ -31,32 +31,30 @@ func ReconcileControlPlanes(ctx context.Context, rc *ReconciliationContext, etcd
 		}), nil
 	}
 
-	if !rc.LBHealthy() {
-		return nil, nil
-	}
-
 	// pending tearing down machines with finalizers should cancel any other operations
 	if len(rc.GetTearingDownMachines()) > 0 {
 		return nil, nil
 	}
 
-	// do a single destroy
-	for _, id := range rc.GetMachinesToTeardown() {
-		clusterMachine, ok := rc.GetClusterMachine(id)
-		if !ok {
-			continue
-		}
+	if rc.LBHealthy() {
+		// do a single destroy
+		for _, id := range rc.GetMachinesToTeardown() {
+			clusterMachine, ok := rc.GetClusterMachine(id)
+			if !ok {
+				continue
+			}
 
-		etcdStatus, err := etcdStatusGetter(ctx)
-		if err != nil {
-			return nil, err
-		}
+			etcdStatus, err := etcdStatusGetter(ctx)
+			if err != nil {
+				return nil, err
+			}
 
-		if err := check.CanScaleDown(etcdStatus, clusterMachine); err != nil {
-			return nil, err
-		}
+			if err := check.CanScaleDown(etcdStatus, clusterMachine); err != nil {
+				return nil, err
+			}
 
-		return []Operation{&Teardown{ID: id}}, nil
+			return []Operation{&Teardown{ID: id}}, nil
+		}
 	}
 
 	// do a single update