Merge branch 'main' into reflect-cluster-status

magnum_cluster_api/driver.py

@@ -31,6 +31,8 @@ def create_cluster(self, context, cluster, cluster_create_timeout):
         cluster.stack_id = utils.generate_cluster_api_name(self.k8s_api)
         cluster.save()
 
+        utils.validate_cluster(cluster)
+
         osc = clients.get_openstack_api(context)
 
         resources.Namespace(self.k8s_api).apply()
@@ -180,6 +182,8 @@ def resize_cluster(
         nodes_to_remove,
         nodegroup=None,
     ):
+        utils.validate_cluster(cluster)
+
         if nodegroup is None:
             nodegroup = cluster.default_ng_worker
 
@@ -252,6 +256,8 @@ def wait_capi_cluster_reconciliation_start(
         raise exceptions.ClusterAPIReconcileTimeout()
 
     def delete_cluster(self, context, cluster):
+        if cluster.stack_id is None:
+            return
         # NOTE(mnaser): This should be removed when this is fixed:
         #
         #               https://github.com/kubernetes-sigs/cluster-api-provider-openstack/issues/842
@@ -320,7 +326,15 @@ def update_nodegroup_status(self, context, cluster, nodegroup):
     def update_nodegroup(self, context, cluster, nodegroup):
         # TODO
 
+        # NOTE(okozachenko1203): First we save the nodegroup status because update_cluster_status()
+        #                        could be finished before update_nodegroup().
+        nodegroup.save()
         resources.apply_cluster_from_magnum_cluster(context, self.k8s_api, cluster)
+        # NOTE(okozachenko1203): We set the cluster status as UPDATE_IN_PROGRESS again at the end because
+        #                        update_cluster_status() could be finished and cluster status has been set as
+        #                        UPDATE_COMPLETE before nodegroup_conductor.Handler.nodegroup_update finished.
+        cluster.status = "UPDATE_IN_PROGRESS"
+        cluster.save()
 
     def delete_nodegroup(self, context, cluster, nodegroup):
         nodegroup.status = "DELETE_IN_PROGRESS"

magnum_cluster_api/exceptions.py

@@ -47,3 +47,7 @@ class OpenStackClusterSubnetNotReady(OpenStackClusterNotReady):
 
 class ClusterAPIReconcileTimeout(Exception):
     pass
+
+
+class ClusterMasterCountEven(Exception):
+    pass

magnum_cluster_api/resources.py

@@ -383,9 +383,23 @@ def get_object(self) -> objects.ClusterResourceSet:
 
 
 class CertificateAuthoritySecret(ClusterBase):
+    def delete(self) -> None:
+        resource = self.get_or_none()
+        if resource:
+            resource.delete()
+
+    def get_or_none(self) -> pykube.Secret:
+        return pykube.Secret.objects(self.api, namespace="magnum-system").get_or_none(
+            name=f"{self.cluster.stack_id}-{self.CERT}"
+        )
+
     def get_object(self) -> pykube.Secret:
+        cert_ref = getattr(self.cluster, self.REF)
+        if cert_ref is None:
+            raise Exception("Certificate for %s doesn't exist." % self.REF)
+
         ca_cert = cert_manager.get_backend().CertManager.get_cert(
-            getattr(self.cluster, self.REF),
+            cert_ref,
             resource_ref=self.cluster.uuid,
         )
 

magnum_cluster_api/utils.py

@@ -27,7 +27,9 @@
 from oslo_utils import strutils
 from tenacity import retry, retry_if_exception_type
 
-from magnum_cluster_api import clients, image_utils, images, objects
+from magnum_cluster_api import clients
+from magnum_cluster_api import exceptions as mcapi_exceptions
+from magnum_cluster_api import image_utils, images, objects
 
 LOG = logging.getLogger(__name__)
 
@@ -316,3 +318,9 @@ def format_event_message(event: pykube.Event):
         event.obj["reason"],
         event.obj["message"],
     )
+
+
+def validate_cluster(cluster: magnum_objects.Cluster):
+    # Check master count
+    if (cluster.master_count % 2) == 0:
+        raise mcapi_exceptions.ClusterMasterCountEven