diff --git a/lib/charms/observability_libs/v0/kubernetes_compute_resources_patch.py b/lib/charms/observability_libs/v0/kubernetes_compute_resources_patch.py index 2ab8a22c..34dd0264 100644 --- a/lib/charms/observability_libs/v0/kubernetes_compute_resources_patch.py +++ b/lib/charms/observability_libs/v0/kubernetes_compute_resources_patch.py @@ -4,7 +4,7 @@ """# KubernetesComputeResourcesPatch Library. This library is designed to enable developers to more simply patch the Kubernetes compute resource -limits and requests created by Juju during the deployment of a sidecar charm. +limits and requests created by Juju during the deployment of a charm. When initialised, this library binds a handler to the parent charm's `config-changed` event. The config-changed event is used because it is guaranteed to fire on startup, on upgrade and on @@ -76,6 +76,17 @@ def _resource_spec_from_config(self) -> ResourceRequirements: return ResourceRequirements(limits=spec, requests=spec) ``` +If you wish to pull the state of the resources patch operation and set the charm unit status based on that patch result, +you can achieve that using `get_status()` function. +```python +class SomeCharm(CharmBase): + def __init__(self, *args): + #... + self.framework.observe(self.on.collect_unit_status, self._on_collect_unit_status) + #... + def _on_collect_unit_status(self, event: CollectStatusEvent): + event.add_status(self.resources_patch.get_status()) +``` Additionally, you may wish to use mocks in your charm's unit testing to ensure that the library does not try to make any API calls, or open any files during testing that are unlikely to be @@ -83,12 +94,14 @@ def _resource_spec_from_config(self) -> ResourceRequirements: ```python # ... +from ops import ActiveStatus @patch.multiple( "charm.KubernetesComputeResourcesPatch", _namespace="test-namespace", _is_patched=lambda *a, **kw: True, is_ready=lambda *a, **kw: True, + get_status=lambda _: ActiveStatus(), ) @patch("lightkube.core.client.GenericSyncClient") def setUp(self, *unused): @@ -105,8 +118,9 @@ def setUp(self, *unused): import logging from decimal import Decimal from math import ceil, floor -from typing import Callable, Dict, List, Optional, Union +from typing import Any, Callable, Dict, List, Optional, Tuple, Union +import tenacity from lightkube import ApiError, Client # pyright: ignore from lightkube.core import exceptions from lightkube.models.apps_v1 import StatefulSetSpec @@ -120,8 +134,10 @@ def setUp(self, *unused): from lightkube.resources.core_v1 import Pod from lightkube.types import PatchType from lightkube.utils.quantity import equals_canonically, parse_quantity +from ops import ActiveStatus, BlockedStatus, WaitingStatus from ops.charm import CharmBase from ops.framework import BoundEvent, EventBase, EventSource, Object, ObjectEvents +from ops.model import StatusBase logger = logging.getLogger(__name__) @@ -133,14 +149,16 @@ def setUp(self, *unused): # Increment this PATCH version before using `charmcraft publish-lib` or reset # to 0 if you are raising the major API version -LIBPATCH = 7 +LIBPATCH = 8 _Decimal = Union[Decimal, float, str, int] # types that are potentially convertible to Decimal def adjust_resource_requirements( - limits: Optional[dict], requests: Optional[dict], adhere_to_requests: bool = True + limits: Optional[Dict[Any, Any]], + requests: Optional[Dict[Any, Any]], + adhere_to_requests: bool = True, ) -> ResourceRequirements: """Adjust resource limits so that `limits` and `requests` are consistent with each other. @@ -289,6 +307,18 @@ def sanitize_resource_spec_dict(spec: Optional[dict]) -> Optional[dict]: return d +def _retry_on_condition(exception): + """Retry if the exception is an ApiError with a status code != 403. + + Returns: a boolean value to indicate whether to retry or not. + """ + if isinstance(exception, ApiError) and str(exception.status.code) != "403": + return True + if isinstance(exception, exceptions.ConfigError) or isinstance(exception, ValueError): + return True + return False + + class K8sResourcePatchFailedEvent(EventBase): """Emitted when patching fails.""" @@ -385,27 +415,132 @@ def get_actual(self, pod_name: str) -> Optional[ResourceRequirements]: ) return podspec.resources + def is_failed( + self, resource_reqs_func: Callable[[], ResourceRequirements] + ) -> Tuple[bool, str]: + """Returns a tuple indicating whether a patch operation has failed along with a failure message. + + Implementation is based on dry running the patch operation to catch if there would be failures (e.g: Wrong spec and Auth errors). + """ + try: + resource_reqs = resource_reqs_func() + limits = resource_reqs.limits + requests = resource_reqs.requests + except ValueError as e: + msg = f"Failed obtaining resource limit spec: {e}" + logger.error(msg) + return True, msg + + # Dry run does not catch negative values for resource requests and limits. + if not is_valid_spec(limits) or not is_valid_spec(requests): + msg = f"Invalid resource requirements specs: {limits}, {requests}" + logger.error(msg) + return True, msg + + resource_reqs = ResourceRequirements( + limits=sanitize_resource_spec_dict(limits), # type: ignore[arg-type] + requests=sanitize_resource_spec_dict(requests), # type: ignore[arg-type] + ) + + try: + self.apply(resource_reqs, dry_run=True) + except ApiError as e: + if e.status.code == 403: + msg = f"Kubernetes resources patch failed: `juju trust` this application. {e}" + else: + msg = f"Kubernetes resources patch failed: {e}" + return True, msg + except ValueError as e: + msg = f"Kubernetes resources patch failed: {e}" + return True, msg + + return False, "" + + def is_in_progress(self) -> bool: + """Returns a boolean to indicate whether a patch operation is in progress. + + Implementation follows a similar approach to `kubectl rollout status statefulset` to track the progress of a rollout. + Reference: https://github.com/kubernetes/kubectl/blob/kubernetes-1.31.0/pkg/polymorphichelpers/rollout_status.go + """ + try: + sts = self.client.get( + StatefulSet, name=self.statefulset_name, namespace=self.namespace + ) + except (ValueError, ApiError) as e: + # Assumption: if there was a persistent issue, it'd have been caught in `is_failed` + # Wait until next run to try again. + logger.error(f"Failed to fetch statefulset from K8s api: {e}") + return False + + if sts.status is None or sts.spec is None: + logger.debug("status/spec are not yet available") + return False + if sts.status.observedGeneration == 0 or ( + sts.metadata + and sts.status.observedGeneration + and sts.metadata.generation + and sts.metadata.generation > sts.status.observedGeneration + ): + logger.debug("waiting for statefulset spec update to be observed...") + return True + if ( + sts.spec.replicas is not None + and sts.status.readyReplicas is not None + and sts.status.readyReplicas < sts.spec.replicas + ): + logger.debug( + f"Waiting for {sts.spec.replicas-sts.status.readyReplicas} pods to be ready..." + ) + return True + + if ( + sts.spec.updateStrategy + and sts.spec.updateStrategy.type == "rollingUpdate" + and sts.spec.updateStrategy.rollingUpdate is not None + ): + if ( + sts.spec.replicas is not None + and sts.spec.updateStrategy.rollingUpdate.partition is not None + ): + if sts.status.updatedReplicas and sts.status.updatedReplicas < ( + sts.spec.replicas - sts.spec.updateStrategy.rollingUpdate.partition + ): + logger.debug( + f"Waiting for partitioned roll out to finish: {sts.status.updatedReplicas} out of {sts.spec.replicas - sts.spec.updateStrategy.rollingUpdate.partition} new pods have been updated..." + ) + return True + logger.debug( + f"partitioned roll out complete: {sts.status.updatedReplicas} new pods have been updated..." + ) + return False + + if sts.status.updateRevision != sts.status.currentRevision: + logger.debug( + f"waiting for statefulset rolling update to complete {sts.status.updatedReplicas} pods at revision {sts.status.updateRevision}..." + ) + return True + + logger.debug( + f"statefulset rolling update complete pods at revision {sts.status.currentRevision}" + ) + return False + def is_ready(self, pod_name, resource_reqs: ResourceRequirements): """Reports if the resource patch has been applied and is in effect. Returns: bool: A boolean indicating if the service patch has been applied and is in effect. """ - logger.info( - "reqs=%s, templated=%s, actual=%s", - resource_reqs, - self.get_templated(), - self.get_actual(pod_name), - ) return self.is_patched(resource_reqs) and equals_canonically( # pyright: ignore resource_reqs, self.get_actual(pod_name) # pyright: ignore ) - def apply(self, resource_reqs: ResourceRequirements) -> None: + def apply(self, resource_reqs: ResourceRequirements, dry_run=False) -> None: """Patch the Kubernetes resources created by Juju to limit cpu or mem.""" # Need to ignore invalid input, otherwise the StatefulSet gives "FailedCreate" and the # charm would be stuck in unknown/lost. - if self.is_patched(resource_reqs): + if not dry_run and self.is_patched(resource_reqs): + logger.debug(f"Resource requests are already patched: {resource_reqs}") return self.client.patch( @@ -415,6 +550,7 @@ def apply(self, resource_reqs: ResourceRequirements) -> None: namespace=self.namespace, patch_type=PatchType.APPLY, field_manager=self.__class__.__name__, + dry_run=dry_run, ) @@ -422,6 +558,9 @@ class KubernetesComputeResourcesPatch(Object): """A utility for patching the Kubernetes compute resources set up by Juju.""" on = K8sResourcePatchEvents() # pyright: ignore + PATCH_RETRY_STOP = tenacity.stop_after_delay(20) + PATCH_RETRY_WAIT = tenacity.wait_fixed(5) + PATCH_RETRY_IF = tenacity.retry_if_exception(_retry_on_condition) def __init__( self, @@ -468,7 +607,11 @@ def _on_config_changed(self, _): self._patch() def _patch(self) -> None: - """Patch the Kubernetes resources created by Juju to limit cpu or mem.""" + """Patch the Kubernetes resources created by Juju to limit cpu or mem. + + This method will keep on retrying to patch the kubernetes resource for a default duration of 20 seconds + if the patching failure is due to a recoverable error (e.g: Network Latency). + """ try: resource_reqs = self.resource_reqs_func() limits = resource_reqs.limits @@ -492,7 +635,18 @@ def _patch(self) -> None: ) try: - self.patcher.apply(resource_reqs) + for attempt in tenacity.Retrying( + retry=self.PATCH_RETRY_IF, + stop=self.PATCH_RETRY_STOP, + wait=self.PATCH_RETRY_WAIT, + # if you don't succeed raise the last caught exception when you're done + reraise=True, + ): + with attempt: + logger.debug( + f"attempt #{attempt.retry_state.attempt_number} to patch resource limits" + ) + self.patcher.apply(resource_reqs) except exceptions.ConfigError as e: msg = f"Error creating k8s client: {e}" @@ -503,6 +657,7 @@ def _patch(self) -> None: except ApiError as e: if e.status.code == 403: msg = f"Kubernetes resources patch failed: `juju trust` this application. {e}" + else: msg = f"Kubernetes resources patch failed: {e}" @@ -554,6 +709,29 @@ def is_ready(self) -> bool: self.on.patch_failed.emit(message=msg) return False + def get_status(self) -> StatusBase: + """Return the status of patching the resource limits in a `StatusBase` format. + + Returns: + StatusBase: There is a 1:1 mapping between the state of the patching operation and a `StatusBase` value that the charm can be set to. + Possible values are: + - ActiveStatus: The patch was applied successfully. + - BlockedStatus: The patch failed and requires a human intervention. + - WaitingStatus: The patch is still in progress. + + Example: + - ActiveStatus("Patch applied successfully") + - BlockedStatus("Failed due to missing permissions") + - WaitingStatus("Patch is in progress") + """ + failed, msg = self.patcher.is_failed(self.resource_reqs_func) + if failed: + return BlockedStatus(msg) + if self.patcher.is_in_progress(): + return WaitingStatus("waiting for resources patch to apply") + # patch successful or nothing has been patched yet + return ActiveStatus() + @property def _app(self) -> str: """Name of the current Juju application.