controller.go

/*
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

    http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/

package garbagecollection

import (
	"context"
	"fmt"
	"time"

	"github.com/awslabs/operatorpkg/singleton"
	"github.com/samber/lo"
	"go.uber.org/multierr"
	corev1 "k8s.io/api/core/v1"
	"k8s.io/apimachinery/pkg/util/sets"
	"k8s.io/client-go/util/workqueue"
	"k8s.io/klog/v2"
	controllerruntime "sigs.k8s.io/controller-runtime"
	"sigs.k8s.io/controller-runtime/pkg/client"
	"sigs.k8s.io/controller-runtime/pkg/log"
	"sigs.k8s.io/controller-runtime/pkg/manager"
	"sigs.k8s.io/controller-runtime/pkg/reconcile"
	karpv1 "sigs.k8s.io/karpenter/pkg/apis/v1"
	"sigs.k8s.io/karpenter/pkg/cloudprovider"
	"sigs.k8s.io/karpenter/pkg/operator/injection"

	"github.com/cloudpilot-ai/karpenter-provider-alibabacloud/pkg/providers/instance"
)

type Controller struct {
	kubeClient    client.Client
	cloudProvider cloudprovider.CloudProvider
	// Keeps track of successful reconciles for more aggressive requeueing near the start of the controller.
	successfulCount uint64
}

func NewController(kubeClient client.Client, cloudProvider cloudprovider.CloudProvider) *Controller {
	return &Controller{
		kubeClient:      kubeClient,
		cloudProvider:   cloudProvider,
		successfulCount: 0,
	}
}

func (c *Controller) Reconcile(ctx context.Context) (reconcile.Result, error) {
	ctx = injection.WithControllerName(ctx, "instance.garbagecollection")

	// We LIST machines on the CloudProvider BEFORE we grab Machines/Nodes on the cluster so that we make sure that, if
	// LISTing instances takes a long time, our information is more updated by the time we get to Machine and Node LIST
	// This works since our CloudProvider instances are deleted based on whether the Machine exists or not, not vise-versa
	retrieved, err := c.cloudProvider.List(ctx)
	if err != nil {
		return reconcile.Result{}, fmt.Errorf("listing cloudprovider machines, %w", err)
	}
	managedRetrieved := lo.Filter(retrieved, func(nc *karpv1.NodeClaim, _ int) bool {
		return nc.DeletionTimestamp.IsZero()
	})
	nodeClaimList := &karpv1.NodeClaimList{}
	if err = c.kubeClient.List(ctx, nodeClaimList); err != nil {
		return reconcile.Result{}, err
	}
	nodeList := &corev1.NodeList{}
	if err = c.kubeClient.List(ctx, nodeList); err != nil {
		return reconcile.Result{}, err
	}
	resolvedProviderIDs := sets.New[string](lo.FilterMap(nodeClaimList.Items, func(n karpv1.NodeClaim, _ int) (string, bool) {
		return n.Status.ProviderID, n.Status.ProviderID != ""
	})...)
	errs := make([]error, len(retrieved))
	workqueue.ParallelizeUntil(ctx, 100, len(managedRetrieved), func(i int) {
		if !resolvedProviderIDs.Has(managedRetrieved[i].Status.ProviderID) &&
			time.Since(managedRetrieved[i].CreationTimestamp.Time) > time.Second*30 {
			errs[i] = c.garbageCollect(ctx, managedRetrieved[i], nodeList)
		}
	})
	if err = multierr.Combine(errs...); err != nil {
		return reconcile.Result{}, err
	}
	c.successfulCount++
	return reconcile.Result{RequeueAfter: lo.Ternary(c.successfulCount <= 20, time.Second*10, time.Minute)}, nil
}

func (c *Controller) garbageCollect(ctx context.Context, nodeClaim *karpv1.NodeClaim, nodeList *corev1.NodeList) error {
	ctx = log.IntoContext(ctx, log.FromContext(ctx).WithValues("provider-id", nodeClaim.Status.ProviderID))
	if err := c.cloudProvider.Delete(ctx, nodeClaim); err != nil {
		// For instanceStateOperationNotSupportedError, not necessary to throw an error, this happens when you scale up
		// one deployment, then scale down immediately. The node is in initializing state, and cloud SDK api will throw
		// an unsupported operation error, this controller will reconcile again
		if instance.IsInstanceStateOperationNotSupportedError(err) ||
			cloudprovider.IsNodeClaimNotFoundError(err) {
			return nil
		}
		return err
	}
	log.FromContext(ctx).V(1).Info("garbage collected cloudprovider instance")

	// Go ahead and cleanup the node if we know that it exists to make scheduling go quicker
	if node, ok := lo.Find(nodeList.Items, func(n corev1.Node) bool {
		return n.Spec.ProviderID == nodeClaim.Status.ProviderID
	}); ok {
		if err := c.kubeClient.Delete(ctx, &node); err != nil {
			return client.IgnoreNotFound(err)
		}
		log.FromContext(ctx).WithValues("Node", klog.KRef("", node.Name)).V(1).Info("garbage collected node")
	}
	return nil
}

func (c *Controller) Register(_ context.Context, m manager.Manager) error {
	return controllerruntime.NewControllerManagedBy(m).
		Named("instance.garbagecollection").
		WatchesRawSource(singleton.Source()).
		Complete(singleton.AsReconciler(c))
}