fix: move to Blocked state when HugePages are not available (#69)

lib/charms/kubernetes_charm_libraries/v0/hugepages_volumes_patch.py

@@ -75,7 +75,7 @@ def _hugepages_volumes_func_from_config(self) -> list[HugePagesVolume]:
 
 # Increment this PATCH version before using `charmcraft publish-lib` or reset
 # to 0 if you are raising the major API version
-LIBPATCH = 2
+LIBPATCH = 3
 
 logger = logging.getLogger(__name__)
 
@@ -215,11 +215,11 @@ def _statefulset_contains_requested_volumes(
         Returns:
             bool: Whether the StatefulSet contains the given volumes.
         """
-        if not statefulset_spec.template.spec.volumes:
+        if not statefulset_spec.template.spec.volumes:  # type: ignore[union-attr]
             return False
         return all(
             [
-                requested_volume in statefulset_spec.template.spec.volumes
+                requested_volume in statefulset_spec.template.spec.volumes  # type: ignore[union-attr]  # noqa E501
                 for requested_volume in requested_volumes
             ]
         )
@@ -243,7 +243,7 @@ def _pod_contains_requested_volumemounts(
         container = self._get_container(container_name=container_name, containers=containers)
         return all(
             [
-                requested_volumemount in container.volumeMounts
+                requested_volumemount in container.volumeMounts  # type: ignore[operator]
                 for requested_volumemount in requested_volumemounts
             ]
         )
@@ -267,15 +267,15 @@ def _pod_resources_are_set(
         container = self._get_container(container_name=container_name, containers=containers)
         if requested_resources.limits:
             for limit, value in requested_resources.limits.items():
-                if not container.resources.limits:
+                if not container.resources.limits:  # type: ignore[union-attr]
                     return False
-                if container.resources.limits.get(limit) != value:
+                if container.resources.limits.get(limit) != value:  # type: ignore[union-attr]
                     return False
         if requested_resources.requests:
             for request, value in requested_resources.requests.items():
-                if not container.resources.requests:
+                if not container.resources.requests:  # type: ignore[union-attr]
                     return False
-                if container.resources.requests.get(request) != value:
+                if container.resources.requests.get(request) != value:  # type: ignore[union-attr]
                     return False
         return True
 
@@ -368,7 +368,7 @@ def list_volumemounts(self, statefulset_name: str, container_name: str) -> list[
             )
         containers: Iterable[Container] = statefulset.spec.template.spec.containers  # type: ignore[attr-defined]  # noqa: E501
         container = self._get_container(container_name=container_name, containers=containers)
-        return container.volumeMounts
+        return container.volumeMounts  # type: ignore[return-value]
 
     def list_container_resources(
         self, statefulset_name: str, container_name: str
@@ -398,7 +398,7 @@ def list_container_resources(
             Container
         ] = statefulset.spec.template.spec.containers  # type: ignore[attr-defined]  # noqa: E501
         container = self._get_container(container_name=container_name, containers=containers)
-        return container.resources
+        return container.resources  # type: ignore[return-value]
 
 
 class KubernetesHugePagesPatchCharmLib(Object):
@@ -672,8 +672,8 @@ def _generate_resource_requirements_to_be_replaced(self) -> ResourceRequirements
             if current_resources.requests
             else {}
         )
-        new_limits = dict(new_limits.items() | additional_resources.limits.items())
-        new_requests = dict(new_requests.items() | additional_resources.requests.items())
+        new_limits = dict(new_limits.items() | additional_resources.limits.items())  # type: ignore[union-attr]  # noqa E501
+        new_requests = dict(new_requests.items() | additional_resources.requests.items())  # type: ignore[union-attr]  # noqa E501
         new_resources = ResourceRequirements(
             limits=new_limits, requests=new_requests, claims=current_resources.claims
         )

lib/charms/kubernetes_charm_libraries/v0/multus.py

@@ -123,7 +123,7 @@ def _on_config_changed(self, event: EventBase):
 
 # Increment this PATCH version before using `charmcraft publish-lib` or reset
 # to 0 if you are raising the major API version
-LIBPATCH = 10
+LIBPATCH = 12
 
 
 logger = logging.getLogger(__name__)
@@ -351,7 +351,7 @@ def patch_statefulset(
                 )
             )
         if privileged:
-            container.securityContext.privileged = True
+            container.securityContext.privileged = True  # type: ignore[union-attr]
         statefulset_delta = StatefulSet(
             spec=StatefulSetSpec(
                 selector=statefulset.spec.selector,  # type: ignore[attr-defined]
@@ -489,12 +489,12 @@ def _pod_is_patched(
             bool
         """
         if not self._annotations_contains_multus_networks(
-            annotations=pod.metadata.annotations,
+            annotations=pod.metadata.annotations,  # type: ignore[arg-type,union-attr]
             network_annotations=network_annotations,
         ):
             return False
         if not self._container_security_context_is_set(
-            containers=pod.spec.containers,
+            containers=pod.spec.containers,  # type: ignore[union-attr]
             container_name=container_name,
             cap_net_admin=cap_net_admin,
             privileged=privileged,
@@ -537,12 +537,29 @@ def _container_security_context_is_set(
         """
         for container in containers:
             if container.name == container_name:
-                if cap_net_admin and "NET_ADMIN" not in container.securityContext.capabilities.add:
+                if cap_net_admin and "NET_ADMIN" not in container.securityContext.capabilities.add:  # type: ignore[operator,union-attr]  # noqa E501
                     return False
-                if privileged and not container.securityContext.privileged:
+                if privileged and not container.securityContext.privileged:  # type: ignore[union-attr]  # noqa E501
                     return False
         return True
 
+    def multus_is_available(self) -> bool:
+        """Check whether Multus is enabled leveraging existence of NAD custom resource.
+
+        Returns:
+            bool: Whether Multus is enabled
+        """
+        try:
+            list(self.client.list(res=NetworkAttachmentDefinition, namespace=self.namespace))
+        except httpx.HTTPStatusError as e:
+            if e.response.status_code == 404:
+                return False
+            else:
+                raise KubernetesMultusError(
+                    "Unexpected outcome when checking for Multus availability"
+                )
+        return True
+
 
 class KubernetesMultusCharmLib(Object):
     """Class to be instantiated by charms requiring Multus networking."""
@@ -726,3 +743,11 @@ def _on_remove(self, event: RemoveEvent) -> None:
     def delete_pod(self) -> None:
         """Delete the pod."""
         self.kubernetes.delete_pod(self._pod)
+
+    def multus_is_available(self) -> bool:
+        """Check whether Multus is enabled leveraging existence of NAD custom resource.
+
+        Returns:
+            bool: Whether Multus is enabled
+        """
+        return self.kubernetes.multus_is_available()

src/charm.py

@@ -29,7 +29,7 @@
 from lightkube.core.client import Client
 from lightkube.models.core_v1 import ServicePort, ServiceSpec
 from lightkube.models.meta_v1 import ObjectMeta
-from lightkube.resources.core_v1 import Service
+from lightkube.resources.core_v1 import Node, Service
 from ops import RemoveEvent
 from ops.charm import CharmBase, CharmEvents
 from ops.framework import EventBase, EventSource
@@ -55,6 +55,7 @@
 PROMETHEUS_PORT = 8080
 PFCP_PORT = 8805
 REQUIRED_CPU_EXTENSIONS = ["avx2", "rdrand"]
+REQUIRED_CPU_EXTENSIONS_HUGEPAGES = ["pdpe1gb"]
 
 # The default field manager set when using kubectl to create resources
 DEFAULT_FIELD_MANAGER = "controller"
@@ -135,6 +136,7 @@ def __init__(self, *args):
             hugepages_volumes_func=self._volumes_request_func_from_config,
             refresh_event=self.on.hugepages_volumes_config_changed,
         )
+        self.framework.observe(self.on.update_status, self._on_config_changed)
         self.framework.observe(self.on.config_changed, self._on_config_changed)
         self.framework.observe(self.on.bessd_pebble_ready, self._on_bessd_pebble_ready)
         self.framework.observe(self.on.config_storage_attached, self._on_bessd_pebble_ready)
@@ -367,6 +369,16 @@ def _on_config_changed(self, event: EventBase):
         """Handler for config changed events."""
         if not self.unit.is_leader():
             return
+        if self._hugepages_is_enabled():
+            if not self._cpu_is_compatible_for_hugepages():
+                raise IncompatibleCPUError(
+                    "\nCPU is not compatible!\n"
+                    "Please use a CPU that has the following capabilities: "
+                    f"{', '.join(REQUIRED_CPU_EXTENSIONS + REQUIRED_CPU_EXTENSIONS_HUGEPAGES)}"
+                )
+            if not self._hugepages_are_available():
+                self.unit.status = BlockedStatus("Not enough HugePages available")
+                return
         if invalid_configs := self._get_invalid_configs():
             self.unit.status = BlockedStatus(
                 f"The following configurations are not valid: {invalid_configs}"
@@ -788,6 +800,25 @@ def _get_cpu_extensions() -> list[str]:
                 del cpu_flags[0]
         return cpu_flags
 
+    def _cpu_is_compatible_for_hugepages(self) -> bool:
+        return all(
+            required_extension in self._get_cpu_extensions()
+            for required_extension in REQUIRED_CPU_EXTENSIONS_HUGEPAGES
+        )
+
+    @staticmethod
+    def _hugepages_are_available() -> bool:
+        """Checks whether HugePages are available in the K8S nodes.
+
+        Returns:
+            bool: Whether HugePages are available in the K8S nodes
+        """
+        client = Client()
+        nodes = client.list(Node)
+        if not nodes:
+            return False
+        return all([node.status.allocatable.get("hugepages-1Gi", "0") >= "2Gi" for node in nodes])  # type: ignore[union-attr]  # noqa E501
+
     def _get_access_nad_config(self) -> Dict[Any, Any]:
         """Get access interface NAD config.
 

tests/unit/test_charm.py

@@ -8,7 +8,7 @@
 from charms.kubernetes_charm_libraries.v0.multus import (  # type: ignore[import]
     NetworkAttachmentDefinition,
 )
-from lightkube.models.core_v1 import ServicePort, ServiceSpec
+from lightkube.models.core_v1 import Node, NodeStatus, ServicePort, ServiceSpec
 from lightkube.models.meta_v1 import ObjectMeta
 from lightkube.resources.core_v1 import Service
 from ops import testing
@@ -770,6 +770,84 @@ def test_when_remove_then_external_service_is_deleted(self, patch_client):
             namespace=self.namespace,
         )
 
+    @patch("charm.check_output")
+    @patch("charm.Client", new=Mock)
+    @patch(f"{HUGEPAGES_LIBRARY_PATH}.KubernetesHugePagesPatchCharmLib.is_patched")
+    def test_given_cpu_not_supporting_required_hugepages_instructions_when_hugepages_enabled_then_incompatiblecpuerror_is_raised(  # noqa: E501
+        self, patch_hugepages_is_patched, patched_check_output
+    ):
+        patch_hugepages_is_patched.return_value = False
+        patched_check_output.return_value = b"Flags: ssse3 fma cx16 rdrand"
+
+        with self.assertRaises(IncompatibleCPUError):
+            self.harness.update_config(key_values={"enable-hugepages": True})
+
+    @patch("charm.check_output")
+    @patch("lightkube.core.client.GenericSyncClient", new=Mock)
+    @patch("lightkube.core.client.Client.list")
+    @patch(f"{HUGEPAGES_LIBRARY_PATH}.KubernetesHugePagesPatchCharmLib.is_patched")
+    def test_given_cpu_supporting_required_hugepages_instructions_when_hugepages_enabled_then_charm_goes_to_waiting_status(  # noqa: E501
+        self, patch_hugepages_is_patched, patch_list, patched_check_output
+    ):
+        patch_hugepages_is_patched.return_value = True
+        patched_check_output.return_value = b"Flags: avx2 ssse3 fma cx16 rdrand pdpe1gb"
+        patch_list.side_effect = [
+            [Node(status=NodeStatus(allocatable={"hugepages-1Gi": "3Gi"}))],
+            [],
+        ]
+
+        self.harness.update_config(key_values={"enable-hugepages": True})
+
+        self.assertEqual(
+            self.harness.model.unit.status,
+            WaitingStatus("Waiting for bessd container to be ready"),
+        )
+
+    @patch("charm.check_output")
+    @patch("lightkube.core.client.GenericSyncClient", new=Mock)
+    @patch("lightkube.core.client.Client.list")
+    @patch(f"{HUGEPAGES_LIBRARY_PATH}.KubernetesHugePagesPatchCharmLib.is_patched")
+    def test_given_cpu_supporting_required_hugepages_instructions_and_not_available_hugepages_when_hugepages_enabled_then_charm_goes_to_blocked_status(  # noqa: E501
+        self, patch_hugepages_is_patched, patch_list, patched_check_output
+    ):
+        patch_hugepages_is_patched.return_value = False
+        patched_check_output.return_value = b"Flags: avx2 ssse3 fma cx16 rdrand pdpe1gb"
+        patch_list.return_value = [Node(status=NodeStatus(allocatable={"hugepages-1Gi": "1Gi"}))]
+
+        self.harness.update_config(key_values={"enable-hugepages": True})
+
+        self.assertEqual(
+            self.harness.model.unit.status, BlockedStatus("Not enough HugePages available")
+        )
+
+    @patch("charm.check_output")
+    @patch("lightkube.core.client.GenericSyncClient", new=Mock)
+    @patch("lightkube.core.client.Client.list")
+    @patch(f"{HUGEPAGES_LIBRARY_PATH}.KubernetesHugePagesPatchCharmLib.is_patched")
+    def test_given_hugepages_not_available_then_hugepages_available_when_update_status_then_charm_goes_to_waiting_status(  # noqa: E501
+        self, patch_hugepages_is_patched, patch_list, patched_check_output
+    ):
+        patch_hugepages_is_patched.return_value = True
+        patched_check_output.return_value = b"Flags: avx2 ssse3 fma cx16 rdrand pdpe1gb"
+        patch_list.side_effect = [
+            [Node(status=NodeStatus(allocatable={"hugepages-1Gi": "1Gi"}))],
+            [Node(status=NodeStatus(allocatable={"hugepages-1Gi": "2Gi"}))],
+            [],
+        ]
+
+        self.harness.update_config(key_values={"enable-hugepages": True})
+
+        self.assertEqual(
+            self.harness.model.unit.status, BlockedStatus("Not enough HugePages available")
+        )
+
+        self.harness.charm.on.update_status.emit()
+
+        self.assertEqual(
+            self.harness.model.unit.status,
+            WaitingStatus("Waiting for bessd container to be ready"),
+        )
+
     @patch(f"{HUGEPAGES_LIBRARY_PATH}.KubernetesHugePagesPatchCharmLib.is_patched")
     def test_given_default_config_when_network_attachment_definitions_from_config_is_called_then_no_interface_mtu_specified_in_nad(  # noqa: E501
         self,