descheduler/pkg/descheduler/descheduler.go

/*
Copyright 2017 The Kubernetes Authors.

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 descheduler

import (
	"context"
	"fmt"
	"math"
	"strconv"
	"time"

	schedulingv1 "k8s.io/api/scheduling/v1"
	"k8s.io/apimachinery/pkg/runtime/schema"

	"go.opentelemetry.io/otel/attribute"
	"go.opentelemetry.io/otel/trace"
	"k8s.io/apimachinery/pkg/api/meta"
	"k8s.io/client-go/discovery"
	"k8s.io/client-go/informers"
	"k8s.io/client-go/tools/events"
	componentbaseconfig "k8s.io/component-base/config"
	"k8s.io/klog/v2"

	v1 "k8s.io/api/core/v1"
	policy "k8s.io/api/policy/v1"
	"k8s.io/apimachinery/pkg/labels"
	"k8s.io/apimachinery/pkg/runtime"
	utilversion "k8s.io/apimachinery/pkg/util/version"
	"k8s.io/apimachinery/pkg/util/wait"
	clientset "k8s.io/client-go/kubernetes"
	fakeclientset "k8s.io/client-go/kubernetes/fake"
	core "k8s.io/client-go/testing"

	"sigs.k8s.io/descheduler/pkg/descheduler/client"
	eutils "sigs.k8s.io/descheduler/pkg/descheduler/evictions/utils"
	nodeutil "sigs.k8s.io/descheduler/pkg/descheduler/node"
	"sigs.k8s.io/descheduler/pkg/tracing"
	"sigs.k8s.io/descheduler/pkg/utils"
	"sigs.k8s.io/descheduler/pkg/version"

	"sigs.k8s.io/descheduler/cmd/descheduler/app/options"
	"sigs.k8s.io/descheduler/metrics"
	"sigs.k8s.io/descheduler/pkg/api"
	"sigs.k8s.io/descheduler/pkg/descheduler/evictions"
	podutil "sigs.k8s.io/descheduler/pkg/descheduler/pod"
	"sigs.k8s.io/descheduler/pkg/framework/pluginregistry"
	frameworkprofile "sigs.k8s.io/descheduler/pkg/framework/profile"
	frameworktypes "sigs.k8s.io/descheduler/pkg/framework/types"
)

type eprunner func(ctx context.Context, nodes []*v1.Node) *frameworktypes.Status

type profileRunner struct {
	name                      string
	descheduleEPs, balanceEPs eprunner
}

type descheduler struct {
	rs                     *options.DeschedulerServer
	ir                     *informerResources
	getPodsAssignedToNode  podutil.GetPodsAssignedToNodeFunc
	sharedInformerFactory  informers.SharedInformerFactory
	deschedulerPolicy      *api.DeschedulerPolicy
	eventRecorder          events.EventRecorder
	podEvictor             *evictions.PodEvictor
	podEvictionReactionFnc func(*fakeclientset.Clientset) func(action core.Action) (bool, runtime.Object, error)
}

type informerResources struct {
	sharedInformerFactory informers.SharedInformerFactory
	resourceToInformer    map[schema.GroupVersionResource]informers.GenericInformer
}

func newInformerResources(sharedInformerFactory informers.SharedInformerFactory) *informerResources {
	return &informerResources{
		sharedInformerFactory: sharedInformerFactory,
		resourceToInformer:    make(map[schema.GroupVersionResource]informers.GenericInformer),
	}
}

func (ir *informerResources) Uses(resources ...schema.GroupVersionResource) error {
	for _, resource := range resources {
		informer, err := ir.sharedInformerFactory.ForResource(resource)
		if err != nil {
			return err
		}

		ir.resourceToInformer[resource] = informer
	}
	return nil
}

// CopyTo Copy informer subscriptions to the new factory and objects to the fake client so that the backing caches are populated for when listers are used.
func (ir *informerResources) CopyTo(fakeClient *fakeclientset.Clientset, newFactory informers.SharedInformerFactory) error {
	for resource, informer := range ir.resourceToInformer {
		_, err := newFactory.ForResource(resource)
		if err != nil {
			return fmt.Errorf("error getting resource %s: %w", resource, err)
		}

		objects, err := informer.Lister().List(labels.Everything())
		if err != nil {
			return fmt.Errorf("error listing %s: %w", informer, err)
		}

		for _, object := range objects {
			fakeClient.Tracker().Add(object)
		}
	}
	return nil
}

func newDescheduler(rs *options.DeschedulerServer, deschedulerPolicy *api.DeschedulerPolicy, evictionPolicyGroupVersion string, eventRecorder events.EventRecorder, sharedInformerFactory informers.SharedInformerFactory) (*descheduler, error) {
	podInformer := sharedInformerFactory.Core().V1().Pods().Informer()

	ir := newInformerResources(sharedInformerFactory)
	ir.Uses(v1.SchemeGroupVersion.WithResource("pods"),
		v1.SchemeGroupVersion.WithResource("nodes"),
		// Future work could be to let each plugin declare what type of resources it needs; that way dry runs would stay
		// consistent with the real runs without having to keep the list here in sync.
		v1.SchemeGroupVersion.WithResource("namespaces"),                // Used by the defaultevictor plugin
		schedulingv1.SchemeGroupVersion.WithResource("priorityclasses")) // Used by the defaultevictor plugin

	getPodsAssignedToNode, err := podutil.BuildGetPodsAssignedToNodeFunc(podInformer)
	if err != nil {
		return nil, fmt.Errorf("build get pods assigned to node function error: %v", err)
	}

	podEvictor := evictions.NewPodEvictor(
		nil,
		eventRecorder,
		evictions.NewOptions().
			WithPolicyGroupVersion(evictionPolicyGroupVersion).
			WithMaxPodsToEvictPerNode(deschedulerPolicy.MaxNoOfPodsToEvictPerNode).
			WithMaxPodsToEvictPerNamespace(deschedulerPolicy.MaxNoOfPodsToEvictPerNamespace).
			WithMaxPodsToEvictTotal(deschedulerPolicy.MaxNoOfPodsToEvictTotal).
			WithDryRun(rs.DryRun).
			WithMetricsEnabled(!rs.DisableMetrics),
	)

	return &descheduler{
		rs:                     rs,
		ir:                     ir,
		getPodsAssignedToNode:  getPodsAssignedToNode,
		sharedInformerFactory:  sharedInformerFactory,
		deschedulerPolicy:      deschedulerPolicy,
		eventRecorder:          eventRecorder,
		podEvictor:             podEvictor,
		podEvictionReactionFnc: podEvictionReactionFnc,
	}, nil
}

func (d *descheduler) runDeschedulerLoop(ctx context.Context, nodes []*v1.Node) error {
	var span trace.Span
	ctx, span = tracing.Tracer().Start(ctx, "runDeschedulerLoop")
	defer span.End()
	defer func(loopStartDuration time.Time) {
		metrics.DeschedulerLoopDuration.With(map[string]string{}).Observe(time.Since(loopStartDuration).Seconds())
	}(time.Now())

	// if len is still <= 1 error out
	if len(nodes) <= 1 {
		klog.V(1).InfoS("The cluster size is 0 or 1 meaning eviction causes service disruption or degradation. So aborting..")
		return fmt.Errorf("the cluster size is 0 or 1")
	}

	var client clientset.Interface
	// When the dry mode is enable, collect all the relevant objects (mostly pods) under a fake client.
	// So when evicting pods while running multiple strategies in a row have the cummulative effect
	// as is when evicting pods for real.
	if d.rs.DryRun {
		klog.V(3).Infof("Building a cached client from the cluster for the dry run")
		// Create a new cache so we start from scratch without any leftovers
		fakeClient := fakeclientset.NewSimpleClientset()
		// simulate a pod eviction by deleting a pod
		fakeClient.PrependReactor("create", "pods", d.podEvictionReactionFnc(fakeClient))
		fakeSharedInformerFactory := informers.NewSharedInformerFactory(fakeClient, 0)

		err := d.ir.CopyTo(fakeClient, fakeSharedInformerFactory)
		if err != nil {
			return err
		}

		// create a new instance of the shared informer factor from the cached client
		// register the pod informer, otherwise it will not get running
		d.getPodsAssignedToNode, err = podutil.BuildGetPodsAssignedToNodeFunc(fakeSharedInformerFactory.Core().V1().Pods().Informer())
		if err != nil {
			return fmt.Errorf("build get pods assigned to node function error: %v", err)
		}

		fakeCtx, cncl := context.WithCancel(context.TODO())
		defer cncl()
		fakeSharedInformerFactory.Start(fakeCtx.Done())
		fakeSharedInformerFactory.WaitForCacheSync(fakeCtx.Done())

		client = fakeClient
		d.sharedInformerFactory = fakeSharedInformerFactory
	} else {
		client = d.rs.Client
	}

	klog.V(3).Infof("Setting up the pod evictor")
	d.podEvictor.SetClient(client)
	d.podEvictor.ResetCounters()

	d.runProfiles(ctx, client, nodes)

	klog.V(1).InfoS("Number of evicted pods", "totalEvicted", d.podEvictor.TotalEvicted())

	return nil
}

// runProfiles runs all the deschedule plugins of all profiles and
// later runs through all balance plugins of all profiles. (All Balance plugins should come after all Deschedule plugins)
// see https://github.com/kubernetes-sigs/descheduler/issues/979
func (d *descheduler) runProfiles(ctx context.Context, client clientset.Interface, nodes []*v1.Node) {
	var span trace.Span
	ctx, span = tracing.Tracer().Start(ctx, "runProfiles")
	defer span.End()
	var profileRunners []profileRunner
	for _, profile := range d.deschedulerPolicy.Profiles {
		currProfile, err := frameworkprofile.NewProfile(
			profile,
			pluginregistry.PluginRegistry,
			frameworkprofile.WithClientSet(client),
			frameworkprofile.WithSharedInformerFactory(d.sharedInformerFactory),
			frameworkprofile.WithPodEvictor(d.podEvictor),
			frameworkprofile.WithGetPodsAssignedToNodeFnc(d.getPodsAssignedToNode),
		)
		if err != nil {
			klog.ErrorS(err, "unable to create a profile", "profile", profile.Name)
			continue
		}
		profileRunners = append(profileRunners, profileRunner{profile.Name, currProfile.RunDeschedulePlugins, currProfile.RunBalancePlugins})
	}

	for _, profileR := range profileRunners {
		// First deschedule
		status := profileR.descheduleEPs(ctx, nodes)
		if status != nil && status.Err != nil {
			span.AddEvent("failed to perform deschedule operations", trace.WithAttributes(attribute.String("err", status.Err.Error()), attribute.String("profile", profileR.name), attribute.String("operation", tracing.DescheduleOperation)))
			klog.ErrorS(status.Err, "running deschedule extension point failed with error", "profile", profileR.name)
			continue
		}
	}

	for _, profileR := range profileRunners {
		// Balance Later
		status := profileR.balanceEPs(ctx, nodes)
		if status != nil && status.Err != nil {
			span.AddEvent("failed to perform balance operations", trace.WithAttributes(attribute.String("err", status.Err.Error()), attribute.String("profile", profileR.name), attribute.String("operation", tracing.BalanceOperation)))
			klog.ErrorS(status.Err, "running balance extension point failed with error", "profile", profileR.name)
			continue
		}
	}
}

func Run(ctx context.Context, rs *options.DeschedulerServer) error {
	var span trace.Span
	ctx, span = tracing.Tracer().Start(ctx, "Run")
	defer span.End()
	metrics.Register()

	clientConnection := rs.ClientConnection
	if rs.KubeconfigFile != "" && clientConnection.Kubeconfig == "" {
		clientConnection.Kubeconfig = rs.KubeconfigFile
	}
	rsclient, eventClient, err := createClients(clientConnection)
	if err != nil {
		return err
	}
	rs.Client = rsclient
	rs.EventClient = eventClient

	deschedulerPolicy, err := LoadPolicyConfig(rs.PolicyConfigFile, rs.Client, pluginregistry.PluginRegistry)
	if err != nil {
		return err
	}
	if deschedulerPolicy == nil {
		return fmt.Errorf("deschedulerPolicy is nil")
	}

	// Add k8s compatibility warnings to logs
	if err := validateVersionCompatibility(rs.Client.Discovery(), version.Get()); err != nil {
		klog.Warning(err.Error())
	}

	evictionPolicyGroupVersion, err := eutils.SupportEviction(rs.Client)
	if err != nil || len(evictionPolicyGroupVersion) == 0 {
		return err
	}

	runFn := func() error {
		return RunDeschedulerStrategies(ctx, rs, deschedulerPolicy, evictionPolicyGroupVersion)
	}

	if rs.LeaderElection.LeaderElect && rs.DeschedulingInterval.Seconds() == 0 {
		span.AddEvent("Validation Failure", trace.WithAttributes(attribute.String("err", "leaderElection must be used with deschedulingInterval")))
		return fmt.Errorf("leaderElection must be used with deschedulingInterval")
	}

	if rs.LeaderElection.LeaderElect && rs.DryRun {
		klog.V(1).Info("Warning: DryRun is set to True. You need to disable it to use Leader Election.")
	}

	if rs.LeaderElection.LeaderElect && !rs.DryRun {
		if err := NewLeaderElection(runFn, rsclient, &rs.LeaderElection, ctx); err != nil {
			span.AddEvent("Leader Election Failure", trace.WithAttributes(attribute.String("err", err.Error())))
			return fmt.Errorf("leaderElection: %w", err)
		}
		return nil
	}

	return runFn()
}

func validateVersionCompatibility(discovery discovery.DiscoveryInterface, deschedulerVersionInfo version.Info) error {
	kubeServerVersionInfo, err := discovery.ServerVersion()
	if err != nil {
		return fmt.Errorf("failed to discover Kubernetes server version: %v", err)
	}

	kubeServerVersion, err := utilversion.ParseSemantic(kubeServerVersionInfo.String())
	if err != nil {
		return fmt.Errorf("failed to parse Kubernetes server version '%s': %v", kubeServerVersionInfo.String(), err)
	}

	deschedulerMinor, err := strconv.ParseFloat(deschedulerVersionInfo.Minor, 64)
	if err != nil {
		return fmt.Errorf("failed to convert Descheduler minor version '%s' to float: %v", deschedulerVersionInfo.Minor, err)
	}

	kubeServerMinor := float64(kubeServerVersion.Minor())
	if math.Abs(deschedulerMinor-kubeServerMinor) > 3 {
		return fmt.Errorf(
			"descheduler version %s.%s may not be supported on your version of Kubernetes %v."+
				"See compatibility docs for more info: https://github.com/kubernetes-sigs/descheduler#compatibility-matrix",
			deschedulerVersionInfo.Major,
			deschedulerVersionInfo.Minor,
			kubeServerVersionInfo.String(),
		)
	}

	return nil
}

func podEvictionReactionFnc(fakeClient *fakeclientset.Clientset) func(action core.Action) (bool, runtime.Object, error) {
	return func(action core.Action) (bool, runtime.Object, error) {
		if action.GetSubresource() == "eviction" {
			createAct, matched := action.(core.CreateActionImpl)
			if !matched {
				return false, nil, fmt.Errorf("unable to convert action to core.CreateActionImpl")
			}
			eviction, matched := createAct.Object.(*policy.Eviction)
			if !matched {
				return false, nil, fmt.Errorf("unable to convert action object into *policy.Eviction")
			}
			if err := fakeClient.Tracker().Delete(action.GetResource(), eviction.GetNamespace(), eviction.GetName()); err != nil {
				return false, nil, fmt.Errorf("unable to delete pod %v/%v: %v", eviction.GetNamespace(), eviction.GetName(), err)
			}
			return true, nil, nil
		}
		// fallback to the default reactor
		return false, nil, nil
	}
}

func RunDeschedulerStrategies(ctx context.Context, rs *options.DeschedulerServer, deschedulerPolicy *api.DeschedulerPolicy, evictionPolicyGroupVersion string) error {
	var span trace.Span
	ctx, span = tracing.Tracer().Start(ctx, "RunDeschedulerStrategies")
	defer span.End()

	sharedInformerFactory := informers.NewSharedInformerFactoryWithOptions(rs.Client, 0, informers.WithTransform(trimManagedFields))

	var nodeSelector string
	if deschedulerPolicy.NodeSelector != nil {
		nodeSelector = *deschedulerPolicy.NodeSelector
	}

	var eventClient clientset.Interface
	if rs.DryRun {
		eventClient = fakeclientset.NewSimpleClientset()
	} else {
		eventClient = rs.Client
	}
	eventBroadcaster, eventRecorder := utils.GetRecorderAndBroadcaster(ctx, eventClient)
	defer eventBroadcaster.Shutdown()

	descheduler, err := newDescheduler(rs, deschedulerPolicy, evictionPolicyGroupVersion, eventRecorder, sharedInformerFactory)
	if err != nil {
		span.AddEvent("Failed to create new descheduler", trace.WithAttributes(attribute.String("err", err.Error())))
		return err
	}
	ctx, cancel := context.WithCancel(ctx)
	defer cancel()

	sharedInformerFactory.Start(ctx.Done())
	sharedInformerFactory.WaitForCacheSync(ctx.Done())

	wait.NonSlidingUntil(func() {
		// A next context is created here intentionally to avoid nesting the spans via context.
		sCtx, sSpan := tracing.Tracer().Start(ctx, "NonSlidingUntil")
		defer sSpan.End()

		nodes, err := nodeutil.ReadyNodes(sCtx, rs.Client, descheduler.sharedInformerFactory.Core().V1().Nodes().Lister(), nodeSelector)
		if err != nil {
			sSpan.AddEvent("Failed to detect ready nodes", trace.WithAttributes(attribute.String("err", err.Error())))
			klog.Error(err)
			cancel()
			return
		}
		err = descheduler.runDeschedulerLoop(sCtx, nodes)
		if err != nil {
			sSpan.AddEvent("Failed to run descheduler loop", trace.WithAttributes(attribute.String("err", err.Error())))
			klog.Error(err)
			cancel()
			return
		}
		// If there was no interval specified, send a signal to the stopChannel to end the wait.Until loop after 1 iteration
		if rs.DeschedulingInterval.Seconds() == 0 {
			cancel()
		}
	}, rs.DeschedulingInterval, ctx.Done())

	return nil
}

func GetPluginConfig(pluginName string, pluginConfigs []api.PluginConfig) (*api.PluginConfig, int) {
	for idx, pluginConfig := range pluginConfigs {
		if pluginConfig.Name == pluginName {
			return &pluginConfig, idx
		}
	}
	return nil, 0
}

func createClients(clientConnection componentbaseconfig.ClientConnectionConfiguration) (clientset.Interface, clientset.Interface, error) {
	kClient, err := client.CreateClient(clientConnection, "descheduler")
	if err != nil {
		return nil, nil, err
	}

	eventClient, err := client.CreateClient(clientConnection, "")
	if err != nil {
		return nil, nil, err
	}

	return kClient, eventClient, nil
}

func trimManagedFields(obj interface{}) (interface{}, error) {
	if accessor, err := meta.Accessor(obj); err == nil {
		accessor.SetManagedFields(nil)
	}
	return obj, nil
}