Sok3t/descheduler
1
0
Fork 0
mirror of https://github.com/kubernetes-sigs/descheduler.git synced 2026-01-26 05:14:13 +01:00
Code Packages Releases Activity

Set up the pod evictor only once

Browse Source 
Currently, the pod evictor is created during each descheduling cycle
to reset the internal counters and the fake client (in case a dry run is
configured). Instead, create the pod evictor once and reset only what's
needed. So later on the pod evictor can be extended with e.g. a cache
keeping the track of eviction requests that are still in progress and
required more than a single descheduling cycle to complete.
This commit is contained in:
Jan Chaloupka
2024-06-23 18:15:44 +02:00
parent 0901cb18bf
commit 75880226c0
3 changed files with 143 additions and 35 deletions
Download Patch File Download Diff File Expand all files Collapse all files

35
pkg/descheduler/descheduler.go
View File

@@ -77,9 +77,9 @@ type descheduler struct {
priorityClassLister schedulingv1.PriorityClassLister
getPodsAssignedToNode podutil.GetPodsAssignedToNodeFunc
sharedInformerFactory informers.SharedInformerFactory
evictionPolicyGroupVersion string
deschedulerPolicy *api.DeschedulerPolicy
eventRecorder events.EventRecorder
podEvictor *evictions.PodEvictor
}
func newDescheduler(rs *options.DeschedulerServer, deschedulerPolicy *api.DeschedulerPolicy, evictionPolicyGroupVersion string, eventRecorder events.EventRecorder, sharedInformerFactory informers.SharedInformerFactory) (*descheduler, error) {
@@ -94,6 +94,16 @@ func newDescheduler(rs *options.DeschedulerServer, deschedulerPolicy *api.Desche
return nil, fmt.Errorf("build get pods assigned to node function error: %v", err)
}
podEvictor := evictions.NewPodEvictor(
nil,
evictionPolicyGroupVersion,
rs.DryRun,
deschedulerPolicy.MaxNoOfPodsToEvictPerNode,
deschedulerPolicy.MaxNoOfPodsToEvictPerNamespace,
!rs.DisableMetrics,
eventRecorder,
)
return &descheduler{
rs: rs,
podLister: podLister,
@@ -102,9 +112,9 @@ func newDescheduler(rs *options.DeschedulerServer, deschedulerPolicy *api.Desche
priorityClassLister: priorityClassLister,
getPodsAssignedToNode: getPodsAssignedToNode,
sharedInformerFactory: sharedInformerFactory,
evictionPolicyGroupVersion: evictionPolicyGroupVersion,
deschedulerPolicy: deschedulerPolicy,
eventRecorder: eventRecorder,
podEvictor: podEvictor,
}, nil
}
@@ -153,20 +163,13 @@ func (d *descheduler) runDeschedulerLoop(ctx context.Context, nodes []*v1.Node)
client = d.rs.Client
}
klog.V(3).Infof("Building a pod evictor")
podEvictor := evictions.NewPodEvictor(
client,
d.evictionPolicyGroupVersion,
d.rs.DryRun,
d.deschedulerPolicy.MaxNoOfPodsToEvictPerNode,
d.deschedulerPolicy.MaxNoOfPodsToEvictPerNamespace,
!d.rs.DisableMetrics,
d.eventRecorder,
)
klog.V(3).Infof("Setting up the pod evictor")
d.podEvictor.SetClient(client)
d.podEvictor.ResetCounters()
d.runProfiles(ctx, client, nodes, podEvictor)
d.runProfiles(ctx, client, nodes)
klog.V(1).InfoS("Number of evicted pods", "totalEvicted", podEvictor.TotalEvicted())
klog.V(1).InfoS("Number of evicted pods", "totalEvicted", d.podEvictor.TotalEvicted())
return nil
}
@@ -174,7 +177,7 @@ func (d *descheduler) runDeschedulerLoop(ctx context.Context, nodes []*v1.Node)
// 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, podEvictor *evictions.PodEvictor) {
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()
@@ -185,7 +188,7 @@ func (d *descheduler) runProfiles(ctx context.Context, client clientset.Interfac
pluginregistry.PluginRegistry,
frameworkprofile.WithClientSet(client),
frameworkprofile.WithSharedInformerFactory(d.sharedInformerFactory),
frameworkprofile.WithPodEvictor(podEvictor),
frameworkprofile.WithPodEvictor(d.podEvictor),
frameworkprofile.WithGetPodsAssignedToNodeFnc(d.getPodsAssignedToNode),
)
if err != nil {

98
pkg/descheduler/descheduler_test.go
View File

@@ -13,15 +13,18 @@ import (
"k8s.io/apimachinery/pkg/runtime"
apiversion "k8s.io/apimachinery/pkg/version"
fakediscovery "k8s.io/client-go/discovery/fake"
"k8s.io/client-go/informers"
fakeclientset "k8s.io/client-go/kubernetes/fake"
core "k8s.io/client-go/testing"
"sigs.k8s.io/descheduler/cmd/descheduler/app/options"
"sigs.k8s.io/descheduler/pkg/api"
"sigs.k8s.io/descheduler/pkg/api/v1alpha1"
nodeutil "sigs.k8s.io/descheduler/pkg/descheduler/node"
"sigs.k8s.io/descheduler/pkg/framework/pluginregistry"
"sigs.k8s.io/descheduler/pkg/framework/plugins/defaultevictor"
"sigs.k8s.io/descheduler/pkg/framework/plugins/removeduplicates"
"sigs.k8s.io/descheduler/pkg/framework/plugins/removepodsviolatingnodetaints"
"sigs.k8s.io/descheduler/pkg/utils"
deschedulerversion "sigs.k8s.io/descheduler/pkg/version"
"sigs.k8s.io/descheduler/test"
)
@@ -174,7 +177,7 @@ func TestDuplicate(t *testing.T) {
}
if len(evictedPods) == 0 {
t.Fatalf("Unable to evict pod, node taint did not get propagated to descheduler strategies %v\n", err)
t.Fatalf("Unable to evict pods\n")
}
}
@@ -323,3 +326,96 @@ func podEvictionReactionFuc(evictedPods *[]string) func(action core.Action) (boo
return false, nil, nil // fallback to the default reactor
}
}
func initPluginRegistry() {
pluginregistry.PluginRegistry = pluginregistry.NewRegistry()
pluginregistry.Register(removeduplicates.PluginName, removeduplicates.New, &removeduplicates.RemoveDuplicates{}, &removeduplicates.RemoveDuplicatesArgs{}, removeduplicates.ValidateRemoveDuplicatesArgs, removeduplicates.SetDefaults_RemoveDuplicatesArgs, pluginregistry.PluginRegistry)
pluginregistry.Register(defaultevictor.PluginName, defaultevictor.New, &defaultevictor.DefaultEvictor{}, &defaultevictor.DefaultEvictorArgs{}, defaultevictor.ValidateDefaultEvictorArgs, defaultevictor.SetDefaults_DefaultEvictorArgs, pluginregistry.PluginRegistry)
}
func TestPodEvictorReset(t *testing.T) {
initPluginRegistry()
ctx := context.Background()
node1 := test.BuildTestNode("n1", 2000, 3000, 10, nil)
node2 := test.BuildTestNode("n2", 2000, 3000, 10, nil)
p1 := test.BuildTestPod("p1", 100, 0, node1.Name, nil)
p1.Namespace = "dev"
p2 := test.BuildTestPod("p2", 100, 0, node1.Name, nil)
p2.Namespace = "dev"
p3 := test.BuildTestPod("p3", 100, 0, node1.Name, nil)
p3.Namespace = "dev"
p4 := test.BuildTestPod("p4", 100, 0, node1.Name, nil)
p4.Namespace = "dev"
ownerRef1 := test.GetReplicaSetOwnerRefList()
p1.ObjectMeta.OwnerReferences = ownerRef1
p2.ObjectMeta.OwnerReferences = ownerRef1
p3.ObjectMeta.OwnerReferences = ownerRef1
p4.ObjectMeta.OwnerReferences = ownerRef1
client := fakeclientset.NewSimpleClientset(node1, node2, p1, p2, p3, p4)
eventClient := fakeclientset.NewSimpleClientset(node1, node2, p1, p2, p3, p4)
dp := &v1alpha1.DeschedulerPolicy{
Strategies: v1alpha1.StrategyList{
"RemoveDuplicates": v1alpha1.DeschedulerStrategy{
Enabled: true,
},
},
}
internalDeschedulerPolicy := &api.DeschedulerPolicy{}
scope := scope{}
err := v1alpha1.V1alpha1ToInternal(dp, pluginregistry.PluginRegistry, internalDeschedulerPolicy, scope)
if err != nil {
t.Fatalf("Unable to convert v1alpha1 to v1alpha2: %v", err)
}
rs, err := options.NewDeschedulerServer()
if err != nil {
t.Fatalf("Unable to initialize server: %v", err)
}
rs.Client = client
rs.EventClient = eventClient
var evictedPods []string
client.PrependReactor("create", "pods", podEvictionReactionFuc(&evictedPods))
sharedInformerFactory := informers.NewSharedInformerFactoryWithOptions(rs.Client, 0, informers.WithTransform(trimManagedFields))
eventBroadcaster, eventRecorder := utils.GetRecorderAndBroadcaster(ctx, client)
defer eventBroadcaster.Shutdown()
descheduler, err := newDescheduler(rs, internalDeschedulerPolicy, "v1", eventRecorder, sharedInformerFactory)
if err != nil {
t.Fatalf("Unable to create a descheduler instance: %v", err)
}
ctx, cancel := context.WithCancel(ctx)
defer cancel()
sharedInformerFactory.Start(ctx.Done())
sharedInformerFactory.WaitForCacheSync(ctx.Done())
nodes, err := nodeutil.ReadyNodes(ctx, rs.Client, descheduler.nodeLister, "")
if err != nil {
t.Fatalf("Unable to get ready nodes: %v", err)
}
// a single pod eviction expected
err = descheduler.runDeschedulerLoop(ctx, nodes)
if err != nil {
t.Fatalf("Unable to run a descheduling loop: %v", err)
}
if descheduler.podEvictor.TotalEvicted() != 2 || len(evictedPods) != 2 {
t.Fatalf("Expected (2,2) pods evicted, got (%v, %v) instead", descheduler.podEvictor.TotalEvicted(), len(evictedPods))
}
// a single pod eviction expected
err = descheduler.runDeschedulerLoop(ctx, nodes)
if err != nil {
t.Fatalf("Unable to run a descheduling loop: %v", err)
}
if descheduler.podEvictor.TotalEvicted() != 2 || len(evictedPods) != 4 {
t.Fatalf("Expected (2,4) pods evicted, got (%v, %v) instead", descheduler.podEvictor.TotalEvicted(), len(evictedPods))
}
}

9
pkg/descheduler/evictions/evictions.go
View File

@@ -97,6 +97,15 @@ func (pe *PodEvictor) NodeLimitExceeded(node *v1.Node) bool {
return false
}
func (pe *PodEvictor) ResetCounters() {
pe.nodepodCount = make(nodePodEvictedCount)
pe.namespacePodCount = make(namespacePodEvictCount)
}
func (pe *PodEvictor) SetClient(client clientset.Interface) {
pe.client = client
}
// EvictOptions provides a handle for passing additional info to EvictPod
type EvictOptions struct {
// Reason allows for passing details about the specific eviction for logging.