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"

	v1 "k8s.io/api/core/v1"
	"k8s.io/apimachinery/pkg/util/wait"
	"k8s.io/client-go/informers"
	clientset "k8s.io/client-go/kubernetes"
	"k8s.io/klog/v2"

	"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/client"
	"sigs.k8s.io/descheduler/pkg/descheduler/evictions"
	eutils "sigs.k8s.io/descheduler/pkg/descheduler/evictions/utils"
	nodeutil "sigs.k8s.io/descheduler/pkg/descheduler/node"
	podutil "sigs.k8s.io/descheduler/pkg/descheduler/pod"
	"sigs.k8s.io/descheduler/pkg/descheduler/strategies"
	"sigs.k8s.io/descheduler/pkg/descheduler/strategies/nodeutilization"
)

func Run(rs *options.DeschedulerServer) error {
	metrics.Register()

	ctx := context.Background()
	rsclient, err := client.CreateClient(rs.KubeconfigFile)
	if err != nil {
		return err
	}
	rs.Client = rsclient

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

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

	stopChannel := make(chan struct{})
	return RunDeschedulerStrategies(ctx, rs, deschedulerPolicy, evictionPolicyGroupVersion, stopChannel)
}

type strategyFunction func(ctx context.Context, client clientset.Interface, strategy api.DeschedulerStrategy, nodes []*v1.Node, podEvictor *evictions.PodEvictor, getPodsAssignedToNode podutil.GetPodsAssignedToNodeFunc)

func RunDeschedulerStrategies(ctx context.Context, rs *options.DeschedulerServer, deschedulerPolicy *api.DeschedulerPolicy, evictionPolicyGroupVersion string, stopChannel chan struct{}) error {
	sharedInformerFactory := informers.NewSharedInformerFactory(rs.Client, 0)
	nodeInformer := sharedInformerFactory.Core().V1().Nodes()
	podInformer := sharedInformerFactory.Core().V1().Pods()

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

	sharedInformerFactory.Start(stopChannel)
	sharedInformerFactory.WaitForCacheSync(stopChannel)

	strategyFuncs := map[api.StrategyName]strategyFunction{
		"RemoveDuplicates":                            strategies.RemoveDuplicatePods,
		"LowNodeUtilization":                          nodeutilization.LowNodeUtilization,
		"HighNodeUtilization":                         nodeutilization.HighNodeUtilization,
		"RemovePodsViolatingInterPodAntiAffinity":     strategies.RemovePodsViolatingInterPodAntiAffinity,
		"RemovePodsViolatingNodeAffinity":             strategies.RemovePodsViolatingNodeAffinity,
		"RemovePodsViolatingNodeTaints":               strategies.RemovePodsViolatingNodeTaints,
		"RemovePodsHavingTooManyRestarts":             strategies.RemovePodsHavingTooManyRestarts,
		"PodLifeTime":                                 strategies.PodLifeTime,
		"RemovePodsViolatingTopologySpreadConstraint": strategies.RemovePodsViolatingTopologySpreadConstraint,
		"RemoveFailedPods":                            strategies.RemoveFailedPods,
	}

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

	var evictLocalStoragePods bool
	if deschedulerPolicy.EvictLocalStoragePods != nil {
		evictLocalStoragePods = *deschedulerPolicy.EvictLocalStoragePods
	}

	evictBarePods := false
	if deschedulerPolicy.EvictFailedBarePods != nil {
		evictBarePods = *deschedulerPolicy.EvictFailedBarePods
		if evictBarePods {
			klog.V(1).InfoS("Warning: EvictFailedBarePods is set to True. This could cause eviction of pods without ownerReferences.")
		}
	}

	evictSystemCriticalPods := false
	if deschedulerPolicy.EvictSystemCriticalPods != nil {
		evictSystemCriticalPods = *deschedulerPolicy.EvictSystemCriticalPods
		if evictSystemCriticalPods {
			klog.V(1).InfoS("Warning: EvictSystemCriticalPods is set to True. This could cause eviction of Kubernetes system pods.")
		}
	}

	ignorePvcPods := false
	if deschedulerPolicy.IgnorePVCPods != nil {
		ignorePvcPods = *deschedulerPolicy.IgnorePVCPods
	}

	wait.Until(func() {
		nodes, err := nodeutil.ReadyNodes(ctx, rs.Client, nodeInformer, nodeSelector)
		if err != nil {
			klog.V(1).InfoS("Unable to get ready nodes", "err", err)
			close(stopChannel)
			return
		}

		if len(nodes) <= 1 {
			klog.V(1).InfoS("The cluster size is 0 or 1 meaning eviction causes service disruption or degradation. So aborting..")
			close(stopChannel)
			return
		}

		podEvictor := evictions.NewPodEvictor(
			rs.Client,
			evictionPolicyGroupVersion,
			rs.DryRun,
			deschedulerPolicy.MaxNoOfPodsToEvictPerNode,
			deschedulerPolicy.MaxNoOfPodsToEvictPerNamespace,
			nodes,
			evictLocalStoragePods,
			evictSystemCriticalPods,
			ignorePvcPods,
			evictBarePods,
		)

		for name, strategy := range deschedulerPolicy.Strategies {
			if f, ok := strategyFuncs[name]; ok {
				if strategy.Enabled {
					f(ctx, rs.Client, strategy, nodes, podEvictor, getPodsAssignedToNode)
				}
			} else {
				klog.ErrorS(fmt.Errorf("unknown strategy name"), "skipping strategy", "strategy", name)
			}
		}

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

		// 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 {
			close(stopChannel)
		}
	}, rs.DeschedulingInterval, stopChannel)

	return nil
}