kubernetes
/
autoscaler
mirror of https://github.com/kubernetes/autoscaler.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Remove the MaxNodeProvisioningTimeProvider interface

This commit is contained in:
Bartłomiej Wróblewski 2023-08-03 10:54:48 +00:00
parent e39d1b028d
commit 14655d219f
10 changed files with 73 additions and 212 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

22
cluster-autoscaler/clusterstate/clusterstate.go
View File

@ -28,6 +28,7 @@ import (
"k8s.io/autoscaler/cluster-autoscaler/clusterstate/api"
"k8s.io/autoscaler/cluster-autoscaler/clusterstate/utils"
"k8s.io/autoscaler/cluster-autoscaler/metrics"
"k8s.io/autoscaler/cluster-autoscaler/processors/nodegroupconfig"
"k8s.io/autoscaler/cluster-autoscaler/utils/backoff"
"k8s.io/autoscaler/cluster-autoscaler/utils/gpu"
kube_util "k8s.io/autoscaler/cluster-autoscaler/utils/kubernetes"
@ -50,11 +51,6 @@ var (
errMaxNodeProvisionTimeProviderNotSet = errors.New("MaxNodeProvisionTimeProvider was not set in cluster state")
)
type maxNodeProvisionTimeProvider interface {
// GetMaxNodeProvisionTime returns MaxNodeProvisionTime value that should be used for the given NodeGroup.
GetMaxNodeProvisionTime(nodeGroup cloudprovider.NodeGroup) (time.Duration, error)
}
// ScaleUpRequest contains information about the requested node group scale up.
type ScaleUpRequest struct {
// NodeGroup is the node group to be scaled up.
@ -140,7 +136,7 @@ type ClusterStateRegistry struct {
previousCloudProviderNodeInstances map[string][]cloudprovider.Instance
cloudProviderNodeInstancesCache *utils.CloudProviderNodeInstancesCache
interrupt chan struct{}
maxNodeProvisionTimeProvider maxNodeProvisionTimeProvider
nodeGroupConfigProcessor nodegroupconfig.NodeGroupConfigProcessor
// scaleUpFailures contains information about scale-up failures for each node group. It should be
// cleared periodically to avoid unnecessary accumulation.
@ -148,7 +144,7 @@ type ClusterStateRegistry struct {
}
// NewClusterStateRegistry creates new ClusterStateRegistry.
func NewClusterStateRegistry(cloudProvider cloudprovider.CloudProvider, config ClusterStateRegistryConfig, logRecorder *utils.LogEventRecorder, backoff backoff.Backoff) *ClusterStateRegistry {
func NewClusterStateRegistry(cloudProvider cloudprovider.CloudProvider, config ClusterStateRegistryConfig, logRecorder *utils.LogEventRecorder, backoff backoff.Backoff, nodeGroupConfigProcessor nodegroupconfig.NodeGroupConfigProcessor) *ClusterStateRegistry {
emptyStatus := &api.ClusterAutoscalerStatus{
ClusterwideConditions: make([]api.ClusterAutoscalerCondition, 0),
NodeGroupStatuses: make([]api.NodeGroupStatus, 0),
@ -172,6 +168,7 @@ func NewClusterStateRegistry(cloudProvider cloudprovider.CloudProvider, config C
cloudProviderNodeInstancesCache: utils.NewCloudProviderNodeInstancesCache(cloudProvider),
interrupt: make(chan struct{}),
scaleUpFailures: make(map[string][]ScaleUpFailure),
nodeGroupConfigProcessor: nodeGroupConfigProcessor,
}
}
@ -197,17 +194,10 @@ func (csr *ClusterStateRegistry) RegisterOrUpdateScaleUp(nodeGroup cloudprovider
csr.registerOrUpdateScaleUpNoLock(nodeGroup, delta, currentTime)
}
// RegisterProviders registers providers in the cluster state registry.
func (csr *ClusterStateRegistry) RegisterProviders(maxNodeProvisionTimeProvider maxNodeProvisionTimeProvider) {
csr.maxNodeProvisionTimeProvider = maxNodeProvisionTimeProvider
}
// MaxNodeProvisionTime returns MaxNodeProvisionTime value that should be used for the given NodeGroup.
// TODO(BigDarkClown): remove this method entirely, it is a redundant wrapper
func (csr *ClusterStateRegistry) MaxNodeProvisionTime(nodeGroup cloudprovider.NodeGroup) (time.Duration, error) {
if csr.maxNodeProvisionTimeProvider == nil {
return 0, errMaxNodeProvisionTimeProviderNotSet
}
return csr.maxNodeProvisionTimeProvider.GetMaxNodeProvisionTime(nodeGroup)
return csr.nodeGroupConfigProcessor.GetMaxNodeProvisionTime(nodeGroup)
}
func (csr *ClusterStateRegistry) registerOrUpdateScaleUpNoLock(nodeGroup cloudprovider.NodeGroup, delta int, currentTime time.Time) {

62
cluster-autoscaler/clusterstate/clusterstate_test.go
View File

@ -21,7 +21,9 @@ import (
"testing"
"time"
"k8s.io/autoscaler/cluster-autoscaler/config"
"k8s.io/autoscaler/cluster-autoscaler/metrics"
"k8s.io/autoscaler/cluster-autoscaler/processors/nodegroupconfig"
apiv1 "k8s.io/api/core/v1"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
@ -72,8 +74,7 @@ func TestOKWithScaleUp(t *testing.T) {
clusterstate := NewClusterStateRegistry(provider, ClusterStateRegistryConfig{
MaxTotalUnreadyPercentage: 10,
OkTotalUnreadyCount: 1,
}, fakeLogRecorder, newBackoff())
clusterstate.RegisterProviders(NewMockMaxNodeProvisionTimeProvider(time.Minute))
}, fakeLogRecorder, newBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: time.Minute}))
clusterstate.RegisterOrUpdateScaleUp(provider.GetNodeGroup("ng1"), 4, time.Now())
err := clusterstate.UpdateNodes([]*apiv1.Node{ng1_1, ng2_1}, nil, now)
assert.NoError(t, err)
@ -114,8 +115,7 @@ func TestEmptyOK(t *testing.T) {
clusterstate := NewClusterStateRegistry(provider, ClusterStateRegistryConfig{
MaxTotalUnreadyPercentage: 10,
OkTotalUnreadyCount: 1,
}, fakeLogRecorder, newBackoff())
clusterstate.RegisterProviders(NewMockMaxNodeProvisionTimeProvider(time.Minute))
}, fakeLogRecorder, newBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: time.Minute}))
err := clusterstate.UpdateNodes([]*apiv1.Node{}, nil, now.Add(-5*time.Second))
assert.NoError(t, err)
assert.True(t, clusterstate.IsClusterHealthy())
@ -155,8 +155,7 @@ func TestOKOneUnreadyNode(t *testing.T) {
clusterstate := NewClusterStateRegistry(provider, ClusterStateRegistryConfig{
MaxTotalUnreadyPercentage: 10,
OkTotalUnreadyCount: 1,
}, fakeLogRecorder, newBackoff())
clusterstate.RegisterProviders(NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
}, fakeLogRecorder, newBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 15 * time.Minute}))
err := clusterstate.UpdateNodes([]*apiv1.Node{ng1_1, ng2_1}, nil, now)
assert.NoError(t, err)
assert.True(t, clusterstate.IsClusterHealthy())
@ -194,8 +193,7 @@ func TestNodeWithoutNodeGroupDontCrash(t *testing.T) {
clusterstate := NewClusterStateRegistry(provider, ClusterStateRegistryConfig{
MaxTotalUnreadyPercentage: 10,
OkTotalUnreadyCount: 1,
}, fakeLogRecorder, newBackoff())
clusterstate.RegisterProviders(NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
}, fakeLogRecorder, newBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 15 * time.Minute}))
err := clusterstate.UpdateNodes([]*apiv1.Node{noNgNode}, nil, now)
assert.NoError(t, err)
assert.Empty(t, clusterstate.GetScaleUpFailures())
@ -222,8 +220,7 @@ func TestOKOneUnreadyNodeWithScaleDownCandidate(t *testing.T) {
clusterstate := NewClusterStateRegistry(provider, ClusterStateRegistryConfig{
MaxTotalUnreadyPercentage: 10,
OkTotalUnreadyCount: 1,
}, fakeLogRecorder, newBackoff())
clusterstate.RegisterProviders(NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
}, fakeLogRecorder, newBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 15 * time.Minute}))
err := clusterstate.UpdateNodes([]*apiv1.Node{ng1_1, ng2_1}, nil, now)
clusterstate.UpdateScaleDownCandidates([]*apiv1.Node{ng1_1}, now)
@ -288,8 +285,7 @@ func TestMissingNodes(t *testing.T) {
clusterstate := NewClusterStateRegistry(provider, ClusterStateRegistryConfig{
MaxTotalUnreadyPercentage: 10,
OkTotalUnreadyCount: 1,
}, fakeLogRecorder, newBackoff())
clusterstate.RegisterProviders(NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
}, fakeLogRecorder, newBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 15 * time.Minute}))
err := clusterstate.UpdateNodes([]*apiv1.Node{ng1_1, ng2_1}, nil, now)
assert.NoError(t, err)
assert.True(t, clusterstate.IsClusterHealthy())
@ -331,8 +327,7 @@ func TestTooManyUnready(t *testing.T) {
clusterstate := NewClusterStateRegistry(provider, ClusterStateRegistryConfig{
MaxTotalUnreadyPercentage: 10,
OkTotalUnreadyCount: 1,
}, fakeLogRecorder, newBackoff())
clusterstate.RegisterProviders(NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
}, fakeLogRecorder, newBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 15 * time.Minute}))
err := clusterstate.UpdateNodes([]*apiv1.Node{ng1_1, ng2_1}, nil, now)
assert.NoError(t, err)
assert.False(t, clusterstate.IsClusterHealthy())
@ -361,8 +356,7 @@ func TestUnreadyLongAfterCreation(t *testing.T) {
clusterstate := NewClusterStateRegistry(provider, ClusterStateRegistryConfig{
MaxTotalUnreadyPercentage: 10,
OkTotalUnreadyCount: 1,
}, fakeLogRecorder, newBackoff())
clusterstate.RegisterProviders(NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
}, fakeLogRecorder, newBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 15 * time.Minute}))
err := clusterstate.UpdateNodes([]*apiv1.Node{ng1_1, ng2_1}, nil, now)
assert.NoError(t, err)
assert.Equal(t, 1, len(clusterstate.GetClusterReadiness().Unready))
@ -394,8 +388,7 @@ func TestNotStarted(t *testing.T) {
clusterstate := NewClusterStateRegistry(provider, ClusterStateRegistryConfig{
MaxTotalUnreadyPercentage: 10,
OkTotalUnreadyCount: 1,
}, fakeLogRecorder, newBackoff())
clusterstate.RegisterProviders(NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
}, fakeLogRecorder, newBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 15 * time.Minute}))
err := clusterstate.UpdateNodes([]*apiv1.Node{ng1_1, ng2_1}, nil, now)
assert.NoError(t, err)
assert.Equal(t, 1, len(clusterstate.GetClusterReadiness().NotStarted))
@ -432,8 +425,7 @@ func TestExpiredScaleUp(t *testing.T) {
clusterstate := NewClusterStateRegistry(provider, ClusterStateRegistryConfig{
MaxTotalUnreadyPercentage: 10,
OkTotalUnreadyCount: 1,
}, fakeLogRecorder, newBackoff())
clusterstate.RegisterProviders(NewMockMaxNodeProvisionTimeProvider(2 * time.Minute))
}, fakeLogRecorder, newBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 2 * time.Minute}))
clusterstate.RegisterOrUpdateScaleUp(provider.GetNodeGroup("ng1"), 4, now.Add(-3*time.Minute))
err := clusterstate.UpdateNodes([]*apiv1.Node{ng1_1}, nil, now)
assert.NoError(t, err)
@ -458,9 +450,7 @@ func TestRegisterScaleDown(t *testing.T) {
clusterstate := NewClusterStateRegistry(provider, ClusterStateRegistryConfig{
MaxTotalUnreadyPercentage: 10,
OkTotalUnreadyCount: 1,
}, fakeLogRecorder, newBackoff())
clusterstate.RegisterProviders(NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
}, fakeLogRecorder, newBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 15 * time.Minute}))
now := time.Now()
clusterstate.RegisterScaleDown(&ScaleDownRequest{
@ -528,8 +518,7 @@ func TestUpcomingNodes(t *testing.T) {
clusterstate := NewClusterStateRegistry(provider, ClusterStateRegistryConfig{
MaxTotalUnreadyPercentage: 10,
OkTotalUnreadyCount: 1,
}, fakeLogRecorder, newBackoff())
clusterstate.RegisterProviders(NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
}, fakeLogRecorder, newBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 15 * time.Minute}))
err := clusterstate.UpdateNodes([]*apiv1.Node{ng1_1, ng2_1, ng3_1, ng4_1, ng5_1, ng5_2}, nil, now)
assert.NoError(t, err)
assert.Empty(t, clusterstate.GetScaleUpFailures())
@ -576,8 +565,7 @@ func TestTaintBasedNodeDeletion(t *testing.T) {
clusterstate := NewClusterStateRegistry(provider, ClusterStateRegistryConfig{
MaxTotalUnreadyPercentage: 10,
OkTotalUnreadyCount: 1,
}, fakeLogRecorder, newBackoff())
clusterstate.RegisterProviders(NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
}, fakeLogRecorder, newBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 15 * time.Minute}))
err := clusterstate.UpdateNodes([]*apiv1.Node{ng1_1, ng1_2}, nil, now)
assert.NoError(t, err)
assert.Empty(t, clusterstate.GetScaleUpFailures())
@ -598,8 +586,7 @@ func TestIncorrectSize(t *testing.T) {
clusterstate := NewClusterStateRegistry(provider, ClusterStateRegistryConfig{
MaxTotalUnreadyPercentage: 10,
OkTotalUnreadyCount: 1,
}, fakeLogRecorder, newBackoff())
clusterstate.RegisterProviders(NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
}, fakeLogRecorder, newBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 15 * time.Minute}))
now := time.Now()
clusterstate.UpdateNodes([]*apiv1.Node{ng1_1}, nil, now.Add(-5*time.Minute))
incorrect := clusterstate.incorrectNodeGroupSizes["ng1"]
@ -635,8 +622,7 @@ func TestUnregisteredNodes(t *testing.T) {
clusterstate := NewClusterStateRegistry(provider, ClusterStateRegistryConfig{
MaxTotalUnreadyPercentage: 10,
OkTotalUnreadyCount: 1,
}, fakeLogRecorder, newBackoff())
clusterstate.RegisterProviders(NewMockMaxNodeProvisionTimeProvider(10 * time.Second))
}, fakeLogRecorder, newBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 10 * time.Second}))
err := clusterstate.UpdateNodes([]*apiv1.Node{ng1_1}, nil, time.Now().Add(-time.Minute))
assert.NoError(t, err)
@ -685,8 +671,7 @@ func TestCloudProviderDeletedNodes(t *testing.T) {
clusterstate := NewClusterStateRegistry(provider, ClusterStateRegistryConfig{
MaxTotalUnreadyPercentage: 10,
OkTotalUnreadyCount: 1,
}, fakeLogRecorder, newBackoff())
clusterstate.RegisterProviders(NewMockMaxNodeProvisionTimeProvider(10 * time.Second))
}, fakeLogRecorder, newBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 10 * time.Second}))
now.Add(time.Minute)
err := clusterstate.UpdateNodes([]*apiv1.Node{ng1_1, ng1_2, noNgNode}, nil, now)
@ -887,8 +872,7 @@ func TestScaleUpBackoff(t *testing.T) {
clusterstate := NewClusterStateRegistry(provider, ClusterStateRegistryConfig{
MaxTotalUnreadyPercentage: 10,
OkTotalUnreadyCount: 1,
}, fakeLogRecorder, newBackoff())
clusterstate.RegisterProviders(NewMockMaxNodeProvisionTimeProvider(120 * time.Second))
}, fakeLogRecorder, newBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 120 * time.Second}))
// After failed scale-up, node group should be still healthy, but should backoff from scale-ups
clusterstate.RegisterOrUpdateScaleUp(provider.GetNodeGroup("ng1"), 1, now.Add(-180*time.Second))
@ -955,8 +939,7 @@ func TestGetClusterSize(t *testing.T) {
clusterstate := NewClusterStateRegistry(provider, ClusterStateRegistryConfig{
MaxTotalUnreadyPercentage: 10,
OkTotalUnreadyCount: 1,
}, fakeLogRecorder, newBackoff())
clusterstate.RegisterProviders(NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
}, fakeLogRecorder, newBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 15 * time.Minute}))
// There are 2 actual nodes in 2 node groups with target sizes of 5 and 1.
clusterstate.UpdateNodes([]*apiv1.Node{ng1_1, ng2_1, notAutoscaledNode}, nil, now)
@ -1003,8 +986,8 @@ func TestUpdateScaleUp(t *testing.T) {
},
fakeLogRecorder,
newBackoff(),
nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 10 * time.Second}),
)
clusterstate.RegisterProviders(NewMockMaxNodeProvisionTimeProvider(10 * time.Second))
clusterstate.RegisterOrUpdateScaleUp(provider.GetNodeGroup("ng1"), 100, now)
assert.Equal(t, clusterstate.scaleUpRequests["ng1"].Increase, 100)
@ -1042,8 +1025,7 @@ func TestScaleUpFailures(t *testing.T) {
fakeClient := &fake.Clientset{}
fakeLogRecorder, _ := utils.NewStatusMapRecorder(fakeClient, "kube-system", kube_record.NewFakeRecorder(5), false, "my-cool-configmap")
clusterstate := NewClusterStateRegistry(provider, ClusterStateRegistryConfig{}, fakeLogRecorder, newBackoff())
clusterstate.RegisterProviders(NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
clusterstate := NewClusterStateRegistry(provider, ClusterStateRegistryConfig{}, fakeLogRecorder, newBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 15 * time.Minute}))
clusterstate.RegisterFailedScaleUp(provider.GetNodeGroup("ng1"), metrics.Timeout, "", "", now)
clusterstate.RegisterFailedScaleUp(provider.GetNodeGroup("ng2"), metrics.Timeout, "", "", now)

40
cluster-autoscaler/clusterstate/providers/max_node_provision_time_provider.go
View File

@ -1,40 +0,0 @@
/*
Copyright 2016 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 providers
import (
"time"
"k8s.io/autoscaler/cluster-autoscaler/cloudprovider"
"k8s.io/autoscaler/cluster-autoscaler/context"
"k8s.io/autoscaler/cluster-autoscaler/processors/nodegroupconfig"
)
// NewDefaultMaxNodeProvisionTimeProvider returns the default maxNodeProvisionTimeProvider which uses the NodeGroupConfigProcessor.
func NewDefaultMaxNodeProvisionTimeProvider(context *context.AutoscalingContext, nodeGroupConfigProcessor nodegroupconfig.NodeGroupConfigProcessor) *defultMaxNodeProvisionTimeProvider {
return &defultMaxNodeProvisionTimeProvider{context: context, nodeGroupConfigProcessor: nodeGroupConfigProcessor}
}
type defultMaxNodeProvisionTimeProvider struct {
context *context.AutoscalingContext
nodeGroupConfigProcessor nodegroupconfig.NodeGroupConfigProcessor
}
// GetMaxNodeProvisionTime returns MaxNodeProvisionTime value that should be used for the given NodeGroup.
func (p *defultMaxNodeProvisionTimeProvider) GetMaxNodeProvisionTime(nodeGroup cloudprovider.NodeGroup) (time.Duration, error) {
return p.nodeGroupConfigProcessor.GetMaxNodeProvisionTime(nodeGroup)
}

37
cluster-autoscaler/clusterstate/testutils.go
View File

@ -1,37 +0,0 @@
/*
Copyright 2023 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 clusterstate
import (
"time"
"k8s.io/autoscaler/cluster-autoscaler/cloudprovider"
)
// NewMockMaxNodeProvisionTimeProvider returns static maxNodeProvisionTimeProvider which returns constant MaxNodeProvisionTime for every NodeGroup.
func NewMockMaxNodeProvisionTimeProvider(maxNodeProvisionTime time.Duration) *staticMockMaxNodeProvisionTimeProvider {
return &staticMockMaxNodeProvisionTimeProvider{maxNodeProvisionTime}
}
type staticMockMaxNodeProvisionTimeProvider struct {
staticMaxNodeProvisionTime time.Duration
}
// GetMaxNodeProvisionTime returns constant MaxNodeProvisionTime value that should be used for every NodeGroup.
func (p *staticMockMaxNodeProvisionTimeProvider) GetMaxNodeProvisionTime(cloudprovider.NodeGroup) (time.Duration, error) {
return p.staticMaxNodeProvisionTime, nil
}

6
cluster-autoscaler/core/scaledown/actuation/actuator_test.go
View File

@ -948,8 +948,7 @@ func TestStartDeletion(t *testing.T) {
if err != nil {
t.Fatalf("Couldn't set up autoscaling context: %v", err)
}
csr := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, ctx.LogRecorder, NewBackoff())
csr.RegisterProviders(clusterstate.NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
csr := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, ctx.LogRecorder, NewBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 15 * time.Minute}))
for _, bucket := range tc.emptyNodes {
for _, node := range bucket.Nodes {
err := ctx.ClusterSnapshot.AddNodeWithPods(node, tc.pods[node.Name])
@ -1207,8 +1206,7 @@ func TestStartDeletionInBatchBasic(t *testing.T) {
if err != nil {
t.Fatalf("Couldn't set up autoscaling context: %v", err)
}
csr := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, ctx.LogRecorder, NewBackoff())
csr.RegisterProviders(clusterstate.NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
csr := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, ctx.LogRecorder, NewBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 15 * time.Minute}))
ndt := deletiontracker.NewNodeDeletionTracker(0)
ndb := NewNodeDeletionBatcher(&ctx, csr, ndt, deleteInterval)
evictor := Evictor{EvictionRetryTime: 0, DsEvictionRetryTime: 0, DsEvictionEmptyNodeTimeout: 0, PodEvictionHeadroom: DefaultPodEvictionHeadroom}

4
cluster-autoscaler/core/scaledown/actuation/delete_in_batch_test.go
View File

@ -29,6 +29,7 @@ import (
"k8s.io/autoscaler/cluster-autoscaler/config"
"k8s.io/autoscaler/cluster-autoscaler/core/scaledown/deletiontracker"
. "k8s.io/autoscaler/cluster-autoscaler/core/test"
"k8s.io/autoscaler/cluster-autoscaler/processors/nodegroupconfig"
"k8s.io/autoscaler/cluster-autoscaler/utils/taints"
"k8s.io/client-go/kubernetes/fake"
core "k8s.io/client-go/testing"
@ -162,8 +163,7 @@ func TestRemove(t *testing.T) {
})
ctx, err := NewScaleTestAutoscalingContext(config.AutoscalingOptions{}, fakeClient, nil, provider, nil, nil)
clusterStateRegistry := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, fakeLogRecorder, NewBackoff())
clusterStateRegistry.RegisterProviders(clusterstate.NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
clusterStateRegistry := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, fakeLogRecorder, NewBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 15 * time.Minute}))
if err != nil {
t.Fatalf("Couldn't set up autoscaling context: %v", err)
}

34
cluster-autoscaler/core/scaledown/legacy/legacy_test.go
View File

@ -22,6 +22,7 @@ import (
"testing"
"time"
"k8s.io/autoscaler/cluster-autoscaler/processors/nodegroupconfig"
"k8s.io/autoscaler/cluster-autoscaler/simulator"
"k8s.io/autoscaler/cluster-autoscaler/simulator/clustersnapshot"
autoscaler_errors "k8s.io/autoscaler/cluster-autoscaler/utils/errors"
@ -146,8 +147,7 @@ func TestFindUnneededNodes(t *testing.T) {
context, err := NewScaleTestAutoscalingContext(options, &fake.Clientset{}, registry, provider, nil, nil)
assert.NoError(t, err)
clusterStateRegistry := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff())
clusterStateRegistry.RegisterProviders(clusterstate.NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
clusterStateRegistry := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 15 * time.Minute}))
wrapper := newWrapperForTesting(&context, clusterStateRegistry, nil)
sd := wrapper.sd
allNodes := []*apiv1.Node{n1, n2, n3, n4, n5, n7, n8, n9}
@ -278,8 +278,7 @@ func TestFindUnneededGPUNodes(t *testing.T) {
context, err := NewScaleTestAutoscalingContext(options, &fake.Clientset{}, registry, provider, nil, nil)
assert.NoError(t, err)
clusterStateRegistry := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff())
clusterStateRegistry.RegisterProviders(clusterstate.NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
clusterStateRegistry := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 15 * time.Minute}))
wrapper := newWrapperForTesting(&context, clusterStateRegistry, nil)
sd := wrapper.sd
allNodes := []*apiv1.Node{n1, n2, n3}
@ -394,8 +393,7 @@ func TestFindUnneededWithPerNodeGroupThresholds(t *testing.T) {
context, err := NewScaleTestAutoscalingContext(globalOptions, &fake.Clientset{}, registry, provider, nil, nil)
assert.NoError(t, err)
clusterStateRegistry := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff())
clusterStateRegistry.RegisterProviders(clusterstate.NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
clusterStateRegistry := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 15 * time.Minute}))
wrapper := newWrapperForTesting(&context, clusterStateRegistry, nil)
sd := wrapper.sd
clustersnapshot.InitializeClusterSnapshotOrDie(t, context.ClusterSnapshot, allNodes, allPods)
@ -478,8 +476,7 @@ func TestPodsWithPreemptionsFindUnneededNodes(t *testing.T) {
context, err := NewScaleTestAutoscalingContext(options, &fake.Clientset{}, registry, provider, nil, nil)
assert.NoError(t, err)
clusterStateRegistry := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff())
clusterStateRegistry.RegisterProviders(clusterstate.NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
clusterStateRegistry := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 15 * time.Minute}))
wrapper := newWrapperForTesting(&context, clusterStateRegistry, nil)
sd := wrapper.sd
@ -543,8 +540,7 @@ func TestFindUnneededMaxCandidates(t *testing.T) {
context, err := NewScaleTestAutoscalingContext(options, &fake.Clientset{}, registry, provider, nil, nil)
assert.NoError(t, err)
clusterStateRegistry := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff())
clusterStateRegistry.RegisterProviders(clusterstate.NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
clusterStateRegistry := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 15 * time.Minute}))
wrapper := newWrapperForTesting(&context, clusterStateRegistry, nil)
sd := wrapper.sd
@ -628,8 +624,7 @@ func TestFindUnneededEmptyNodes(t *testing.T) {
context, err := NewScaleTestAutoscalingContext(options, &fake.Clientset{}, registry, provider, nil, nil)
assert.NoError(t, err)
clusterStateRegistry := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff())
clusterStateRegistry.RegisterProviders(clusterstate.NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
clusterStateRegistry := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 15 * time.Minute}))
wrapper := newWrapperForTesting(&context, clusterStateRegistry, nil)
sd := wrapper.sd
@ -686,8 +681,7 @@ func TestFindUnneededNodePool(t *testing.T) {
context, err := NewScaleTestAutoscalingContext(options, &fake.Clientset{}, registry, provider, nil, nil)
assert.NoError(t, err)
clusterStateRegistry := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff())
clusterStateRegistry.RegisterProviders(clusterstate.NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
clusterStateRegistry := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 15 * time.Minute}))
wrapper := newWrapperForTesting(&context, clusterStateRegistry, nil)
sd := wrapper.sd
clustersnapshot.InitializeClusterSnapshotOrDie(t, context.ClusterSnapshot, nodes, pods)
@ -778,8 +772,7 @@ func TestScaleDown(t *testing.T) {
assert.NoError(t, err)
nodes := []*apiv1.Node{n1, n2}
clusterStateRegistry := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff())
clusterStateRegistry.RegisterProviders(clusterstate.NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
clusterStateRegistry := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 15 * time.Minute}))
wrapper := newWrapperForTesting(&context, clusterStateRegistry, nil)
clustersnapshot.InitializeClusterSnapshotOrDie(t, context.ClusterSnapshot, nodes, []*apiv1.Pod{p1, p2})
autoscalererr = wrapper.UpdateClusterState(nodes, nodes, nil, time.Now().Add(-5*time.Minute))
@ -1036,8 +1029,7 @@ func simpleScaleDownEmpty(t *testing.T, config *ScaleTestConfig) {
context, err := NewScaleTestAutoscalingContext(config.Options, fakeClient, registry, provider, nil, nil)
assert.NoError(t, err)
clusterStateRegistry := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff())
clusterStateRegistry.RegisterProviders(clusterstate.NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
clusterStateRegistry := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.Options.NodeGroupDefaults))
wrapper := newWrapperForTesting(&context, clusterStateRegistry, config.NodeDeletionTracker)
clustersnapshot.InitializeClusterSnapshotOrDie(t, context.ClusterSnapshot, nodes, []*apiv1.Pod{})
autoscalererr = wrapper.UpdateClusterState(nodes, nodes, nil, time.Now().Add(-5*time.Minute))
@ -1132,8 +1124,7 @@ func TestNoScaleDownUnready(t *testing.T) {
nodes := []*apiv1.Node{n1, n2}
// N1 is unready so it requires a bigger unneeded time.
clusterStateRegistry := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff())
clusterStateRegistry.RegisterProviders(clusterstate.NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
clusterStateRegistry := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 15 * time.Minute}))
wrapper := newWrapperForTesting(&context, clusterStateRegistry, nil)
clustersnapshot.InitializeClusterSnapshotOrDie(t, context.ClusterSnapshot, nodes, []*apiv1.Pod{p2})
autoscalererr = wrapper.UpdateClusterState(nodes, nodes, nil, time.Now().Add(-5*time.Minute))
@ -1247,8 +1238,7 @@ func TestScaleDownNoMove(t *testing.T) {
nodes := []*apiv1.Node{n1, n2}
clusterStateRegistry := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff())
clusterStateRegistry.RegisterProviders(clusterstate.NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
clusterStateRegistry := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 15 * time.Minute}))
wrapper := newWrapperForTesting(&context, clusterStateRegistry, nil)
clustersnapshot.InitializeClusterSnapshotOrDie(t, context.ClusterSnapshot, nodes, []*apiv1.Pod{p1, p2})
autoscalererr = wrapper.UpdateClusterState(nodes, nodes, nil, time.Now().Add(-5*time.Minute))

29
cluster-autoscaler/core/scaleup/orchestrator/orchestrator_test.go
View File

@ -26,6 +26,7 @@ import (
"testing"
"time"
"k8s.io/autoscaler/cluster-autoscaler/processors/nodegroupconfig"
kube_record "k8s.io/client-go/tools/record"
"k8s.io/component-base/metrics/legacyregistry"
@ -969,9 +970,7 @@ func runSimpleScaleUpTest(t *testing.T, config *ScaleUpTestConfig) *ScaleUpTestR
nodeInfos, err := nodeinfosprovider.NewDefaultTemplateNodeInfoProvider(nil, false).
Process(&context, nodes, []*appsv1.DaemonSet{}, taints.TaintConfig{}, now)
assert.NoError(t, err)
clusterState := clusterstate.
NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff())
clusterState.RegisterProviders(clusterstate.NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
clusterState := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(options.NodeGroupDefaults))
clusterState.UpdateNodes(nodes, nodeInfos, time.Now())
processors := NewTestProcessors(&context)
orchestrator := New()
@ -1072,8 +1071,7 @@ func TestScaleUpUnhealthy(t *testing.T) {
nodes := []*apiv1.Node{n1, n2}
nodeInfos, _ := nodeinfosprovider.NewDefaultTemplateNodeInfoProvider(nil, false).Process(&context, nodes, []*appsv1.DaemonSet{}, taints.TaintConfig{}, now)
clusterState := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff())
clusterState.RegisterProviders(clusterstate.NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
clusterState := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 15 * time.Minute}))
clusterState.UpdateNodes(nodes, nodeInfos, time.Now())
p3 := BuildTestPod("p-new", 550, 0)
@ -1118,8 +1116,7 @@ func TestBinpackingLimiter(t *testing.T) {
Process(&context, nodes, []*appsv1.DaemonSet{}, taints.TaintConfig{}, now)
assert.NoError(t, err)
clusterState := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff())
clusterState.RegisterProviders(clusterstate.NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
clusterState := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 15 * time.Minute}))
clusterState.UpdateNodes(nodes, nodeInfos, time.Now())
extraPod := BuildTestPod("p-new", 500, 0)
@ -1174,8 +1171,7 @@ func TestScaleUpNoHelp(t *testing.T) {
nodes := []*apiv1.Node{n1}
nodeInfos, _ := nodeinfosprovider.NewDefaultTemplateNodeInfoProvider(nil, false).Process(&context, nodes, []*appsv1.DaemonSet{}, taints.TaintConfig{}, now)
clusterState := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff())
clusterState.RegisterProviders(clusterstate.NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
clusterState := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 15 * time.Minute}))
clusterState.UpdateNodes(nodes, nodeInfos, time.Now())
p3 := BuildTestPod("p-new", 500, 0)
@ -1329,8 +1325,7 @@ func TestComputeSimilarNodeGroups(t *testing.T) {
assert.NoError(t, err)
nodeInfos, _ := nodeinfosprovider.NewDefaultTemplateNodeInfoProvider(nil, false).Process(&ctx, nodes, []*appsv1.DaemonSet{}, taints.TaintConfig{}, now)
clusterState := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, ctx.LogRecorder, NewBackoff())
clusterState.RegisterProviders(clusterstate.NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
clusterState := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, ctx.LogRecorder, NewBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 15 * time.Minute}))
assert.NoError(t, clusterState.UpdateNodes(nodes, nodeInfos, time.Now()))
suOrchestrator := &ScaleUpOrchestrator{}
@ -1394,8 +1389,7 @@ func TestScaleUpBalanceGroups(t *testing.T) {
assert.NoError(t, err)
nodeInfos, _ := nodeinfosprovider.NewDefaultTemplateNodeInfoProvider(nil, false).Process(&context, nodes, []*appsv1.DaemonSet{}, taints.TaintConfig{}, now)
clusterState := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff())
clusterState.RegisterProviders(clusterstate.NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
clusterState := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 15 * time.Minute}))
clusterState.UpdateNodes(nodes, nodeInfos, time.Now())
pods := make([]*apiv1.Pod, 0)
@ -1457,8 +1451,7 @@ func TestScaleUpAutoprovisionedNodeGroup(t *testing.T) {
context, err := NewScaleTestAutoscalingContext(options, fakeClient, listers, provider, nil, nil)
assert.NoError(t, err)
clusterState := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff())
clusterState.RegisterProviders(clusterstate.NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
clusterState := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 15 * time.Minute}))
processors := NewTestProcessors(&context)
processors.NodeGroupListProcessor = &MockAutoprovisioningNodeGroupListProcessor{T: t}
@ -1513,8 +1506,7 @@ func TestScaleUpBalanceAutoprovisionedNodeGroups(t *testing.T) {
context, err := NewScaleTestAutoscalingContext(options, fakeClient, listers, provider, nil, nil)
assert.NoError(t, err)
clusterState := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff())
clusterState.RegisterProviders(clusterstate.NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
clusterState := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 15 * time.Minute}))
processors := NewTestProcessors(&context)
processors.NodeGroupListProcessor = &MockAutoprovisioningNodeGroupListProcessor{T: t}
@ -1575,8 +1567,7 @@ func TestScaleUpToMeetNodeGroupMinSize(t *testing.T) {
nodes := []*apiv1.Node{n1, n2}
nodeInfos, _ := nodeinfosprovider.NewDefaultTemplateNodeInfoProvider(nil, false).Process(&context, nodes, []*appsv1.DaemonSet{}, taints.TaintConfig{}, time.Now())
processors := NewTestProcessors(&context)
clusterState := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff())
clusterState.RegisterProviders(clusterstate.NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
clusterState := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{}, context.LogRecorder, NewBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 15 * time.Minute}))
clusterState.UpdateNodes(nodes, nodeInfos, time.Now())
suOrchestrator := New()

4
cluster-autoscaler/core/static_autoscaler.go
View File

@ -35,7 +35,6 @@ import (
"k8s.io/autoscaler/cluster-autoscaler/cloudprovider"
"k8s.io/autoscaler/cluster-autoscaler/clusterstate"
"k8s.io/autoscaler/cluster-autoscaler/clusterstate/providers"
"k8s.io/autoscaler/cluster-autoscaler/clusterstate/utils"
"k8s.io/autoscaler/cluster-autoscaler/config"
"k8s.io/autoscaler/cluster-autoscaler/context"
@ -148,7 +147,7 @@ func NewStaticAutoscaler(
MaxTotalUnreadyPercentage: opts.MaxTotalUnreadyPercentage,
OkTotalUnreadyCount: opts.OkTotalUnreadyCount,
}
clusterStateRegistry := clusterstate.NewClusterStateRegistry(cloudProvider, clusterStateConfig, autoscalingKubeClients.LogRecorder, backoff)
clusterStateRegistry := clusterstate.NewClusterStateRegistry(cloudProvider, clusterStateConfig, autoscalingKubeClients.LogRecorder, backoff, processors.NodeGroupConfigProcessor)
processorCallbacks := newStaticAutoscalerProcessorCallbacks()
autoscalingContext := context.NewAutoscalingContext(
opts,
@ -164,7 +163,6 @@ func NewStaticAutoscaler(
clusterStateRegistry)
taintConfig := taints.NewTaintConfig(opts)
clusterStateRegistry.RegisterProviders(providers.NewDefaultMaxNodeProvisionTimeProvider(autoscalingContext, processors.NodeGroupConfigProcessor))
processors.ScaleDownCandidatesNotifier.Register(clusterStateRegistry)
// TODO: Populate the ScaleDownActuator/Planner fields in AutoscalingContext

47
cluster-autoscaler/core/static_autoscaler_test.go
View File

@ -42,6 +42,7 @@ import (
core_utils "k8s.io/autoscaler/cluster-autoscaler/core/utils"
"k8s.io/autoscaler/cluster-autoscaler/estimator"
ca_processors "k8s.io/autoscaler/cluster-autoscaler/processors"
"k8s.io/autoscaler/cluster-autoscaler/processors/nodegroupconfig"
"k8s.io/autoscaler/cluster-autoscaler/simulator"
"k8s.io/autoscaler/cluster-autoscaler/simulator/clustersnapshot"
"k8s.io/autoscaler/cluster-autoscaler/simulator/utilization"
@ -235,8 +236,7 @@ func TestStaticAutoscalerRunOnce(t *testing.T) {
OkTotalUnreadyCount: 1,
}
processors := NewTestProcessors(&context)
clusterState := clusterstate.NewClusterStateRegistry(provider, clusterStateConfig, context.LogRecorder, NewBackoff())
clusterState.RegisterProviders(clusterstate.NewMockMaxNodeProvisionTimeProvider(options.NodeGroupDefaults.MaxNodeProvisionTime))
clusterState := clusterstate.NewClusterStateRegistry(provider, clusterStateConfig, context.LogRecorder, NewBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(options.NodeGroupDefaults))
sdPlanner, sdActuator := newScaleDownPlannerAndActuator(t, &context, processors, clusterState)
suOrchestrator := orchestrator.New()
suOrchestrator.Initialize(&context, processors, clusterState, taints.TaintConfig{})
@ -453,8 +453,7 @@ func TestStaticAutoscalerRunOnceWithAutoprovisionedEnabled(t *testing.T) {
clusterStateConfig := clusterstate.ClusterStateRegistryConfig{
OkTotalUnreadyCount: 0,
}
clusterState := clusterstate.NewClusterStateRegistry(provider, clusterStateConfig, context.LogRecorder, NewBackoff())
clusterState.RegisterProviders(clusterstate.NewMockMaxNodeProvisionTimeProvider(options.NodeGroupDefaults.MaxNodeProvisionTime))
clusterState := clusterstate.NewClusterStateRegistry(provider, clusterStateConfig, context.LogRecorder, NewBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(options.NodeGroupDefaults))
sdPlanner, sdActuator := newScaleDownPlannerAndActuator(t, &context, processors, clusterState)
suOrchestrator := orchestrator.New()
@ -598,8 +597,7 @@ func TestStaticAutoscalerRunOnceWithALongUnregisteredNode(t *testing.T) {
clusterStateConfig := clusterstate.ClusterStateRegistryConfig{
OkTotalUnreadyCount: 1,
}
clusterState := clusterstate.NewClusterStateRegistry(provider, clusterStateConfig, context.LogRecorder, NewBackoff())
clusterState.RegisterProviders(clusterstate.NewMockMaxNodeProvisionTimeProvider(options.NodeGroupDefaults.MaxNodeProvisionTime))
clusterState := clusterstate.NewClusterStateRegistry(provider, clusterStateConfig, context.LogRecorder, NewBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(options.NodeGroupDefaults))
// broken node detected as unregistered
nodes := []*apiv1.Node{n1}
@ -761,8 +759,7 @@ func TestStaticAutoscalerRunOncePodsWithPriorities(t *testing.T) {
}
processors := NewTestProcessors(&context)
clusterState := clusterstate.NewClusterStateRegistry(provider, clusterStateConfig, context.LogRecorder, NewBackoff())
clusterState.RegisterProviders(clusterstate.NewMockMaxNodeProvisionTimeProvider(options.NodeGroupDefaults.MaxNodeProvisionTime))
clusterState := clusterstate.NewClusterStateRegistry(provider, clusterStateConfig, context.LogRecorder, NewBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(options.NodeGroupDefaults))
sdPlanner, sdActuator := newScaleDownPlannerAndActuator(t, &context, processors, clusterState)
suOrchestrator := orchestrator.New()
suOrchestrator.Initialize(&context, processors, clusterState, taints.TaintConfig{})
@ -897,8 +894,7 @@ func TestStaticAutoscalerRunOnceWithFilteringOnBinPackingEstimator(t *testing.T)
}
processors := NewTestProcessors(&context)
clusterState := clusterstate.NewClusterStateRegistry(provider, clusterStateConfig, context.LogRecorder, NewBackoff())
clusterState.RegisterProviders(clusterstate.NewMockMaxNodeProvisionTimeProvider(options.NodeGroupDefaults.MaxNodeProvisionTime))
clusterState := clusterstate.NewClusterStateRegistry(provider, clusterStateConfig, context.LogRecorder, NewBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(options.NodeGroupDefaults))
sdPlanner, sdActuator := newScaleDownPlannerAndActuator(t, &context, processors, clusterState)
autoscaler := &StaticAutoscaler{
@ -998,8 +994,7 @@ func TestStaticAutoscalerRunOnceWithFilteringOnUpcomingNodesEnabledNoScaleUp(t *
}
processors := NewTestProcessors(&context)
clusterState := clusterstate.NewClusterStateRegistry(provider, clusterStateConfig, context.LogRecorder, NewBackoff())
clusterState.RegisterProviders(clusterstate.NewMockMaxNodeProvisionTimeProvider(options.NodeGroupDefaults.MaxNodeProvisionTime))
clusterState := clusterstate.NewClusterStateRegistry(provider, clusterStateConfig, context.LogRecorder, NewBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(options.NodeGroupDefaults))
sdPlanner, sdActuator := newScaleDownPlannerAndActuator(t, &context, processors, clusterState)
autoscaler := &StaticAutoscaler{
@ -1056,9 +1051,8 @@ func TestStaticAutoscalerInstanceCreationErrors(t *testing.T) {
OkTotalUnreadyCount: 1,
}
staticMaxNodeProvisionTimeProvider := clusterstate.NewMockMaxNodeProvisionTimeProvider(options.NodeGroupDefaults.MaxNodeProvisionTime)
clusterState := clusterstate.NewClusterStateRegistry(provider, clusterStateConfig, context.LogRecorder, NewBackoff())
clusterState.RegisterProviders(staticMaxNodeProvisionTimeProvider)
nodeGroupConfigProcessor := nodegroupconfig.NewDefaultNodeGroupConfigProcessor(options.NodeGroupDefaults)
clusterState := clusterstate.NewClusterStateRegistry(provider, clusterStateConfig, context.LogRecorder, NewBackoff(), nodeGroupConfigProcessor)
autoscaler := &StaticAutoscaler{
AutoscalingContext: &context,
clusterStateRegistry: clusterState,
@ -1296,8 +1290,7 @@ func TestStaticAutoscalerInstanceCreationErrors(t *testing.T) {
return false
}, nil)
clusterState = clusterstate.NewClusterStateRegistry(provider, clusterStateConfig, context.LogRecorder, NewBackoff())
clusterState.RegisterProviders(staticMaxNodeProvisionTimeProvider)
clusterState = clusterstate.NewClusterStateRegistry(provider, clusterStateConfig, context.LogRecorder, NewBackoff(), nodeGroupConfigProcessor)
clusterState.RefreshCloudProviderNodeInstancesCache()
autoscaler.clusterStateRegistry = clusterState
@ -1319,7 +1312,7 @@ func TestStaticAutoscalerInstanceCreationErrors(t *testing.T) {
nodeGroupAtomic.On("GetOptions", options.NodeGroupDefaults).Return(
&config.NodeGroupAutoscalingOptions{
ZeroOrMaxNodeScaling: true,
}, nil).Twice()
}, nil)
nodeGroupAtomic.On("Nodes").Return([]cloudprovider.Instance{
{
Id: "D1",
@ -1354,8 +1347,7 @@ func TestStaticAutoscalerInstanceCreationErrors(t *testing.T) {
return nil
}, nil).Times(3)
clusterState = clusterstate.NewClusterStateRegistry(provider, clusterStateConfig, context.LogRecorder, NewBackoff())
clusterState.RegisterProviders(staticMaxNodeProvisionTimeProvider)
clusterState = clusterstate.NewClusterStateRegistry(provider, clusterStateConfig, context.LogRecorder, NewBackoff(), nodeGroupConfigProcessor)
clusterState.RefreshCloudProviderNodeInstancesCache()
autoscaler.CloudProvider = provider
autoscaler.clusterStateRegistry = clusterState
@ -1498,8 +1490,7 @@ func TestStaticAutoscalerUpcomingScaleDownCandidates(t *testing.T) {
// Create CSR with unhealthy cluster protection effectively disabled, to guarantee we reach the tested logic.
csrConfig := clusterstate.ClusterStateRegistryConfig{OkTotalUnreadyCount: nodeGroupCount * unreadyNodesCount}
csr := clusterstate.NewClusterStateRegistry(provider, csrConfig, ctx.LogRecorder, NewBackoff())
csr.RegisterProviders(clusterstate.NewMockMaxNodeProvisionTimeProvider(15 * time.Minute))
csr := clusterstate.NewClusterStateRegistry(provider, csrConfig, ctx.LogRecorder, NewBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(config.NodeGroupAutoscalingOptions{MaxNodeProvisionTime: 15 * time.Minute}))
// Setting the Actuator is necessary for testing any scale-down logic, it shouldn't have anything to do in this test.
actuator := actuation.NewActuator(&ctx, csr, deletiontracker.NewNodeDeletionTracker(0*time.Second), simulator.NodeDeleteOptions{}, NewTestProcessors(&ctx).NodeGroupConfigProcessor)
@ -1596,8 +1587,7 @@ func TestRemoveFixNodeTargetSize(t *testing.T) {
clusterState := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{
MaxTotalUnreadyPercentage: 10,
OkTotalUnreadyCount: 1,
}, fakeLogRecorder, NewBackoff())
clusterState.RegisterProviders(clusterstate.NewMockMaxNodeProvisionTimeProvider(context.AutoscalingOptions.NodeGroupDefaults.MaxNodeProvisionTime))
}, fakeLogRecorder, NewBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(context.AutoscalingOptions.NodeGroupDefaults))
err := clusterState.UpdateNodes([]*apiv1.Node{ng1_1}, nil, now.Add(-time.Hour))
assert.NoError(t, err)
@ -1645,8 +1635,7 @@ func TestRemoveOldUnregisteredNodes(t *testing.T) {
clusterState := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{
MaxTotalUnreadyPercentage: 10,
OkTotalUnreadyCount: 1,
}, fakeLogRecorder, NewBackoff())
clusterState.RegisterProviders(clusterstate.NewMockMaxNodeProvisionTimeProvider(context.AutoscalingOptions.NodeGroupDefaults.MaxNodeProvisionTime))
}, fakeLogRecorder, NewBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(context.AutoscalingOptions.NodeGroupDefaults))
err := clusterState.UpdateNodes([]*apiv1.Node{ng1_1}, nil, now.Add(-time.Hour))
assert.NoError(t, err)
@ -1680,6 +1669,7 @@ func TestRemoveOldUnregisteredNodesAtomic(t *testing.T) {
return nil
})
provider.AddNodeGroupWithCustomOptions("atomic-ng", 0, 10, 10, &config.NodeGroupAutoscalingOptions{
MaxNodeProvisionTime: 45 * time.Minute,
ZeroOrMaxNodeScaling: true,
})
regNode := BuildTestNode("atomic-ng-0", 1000, 1000)
@ -1697,7 +1687,7 @@ func TestRemoveOldUnregisteredNodesAtomic(t *testing.T) {
context := &context.AutoscalingContext{
AutoscalingOptions: config.AutoscalingOptions{
NodeGroupDefaults: config.NodeGroupAutoscalingOptions{
MaxNodeProvisionTime: 45 * time.Minute,
MaxNodeProvisionTime: time.Hour,
},
},
CloudProvider: provider,
@ -1705,8 +1695,7 @@ func TestRemoveOldUnregisteredNodesAtomic(t *testing.T) {
clusterState := clusterstate.NewClusterStateRegistry(provider, clusterstate.ClusterStateRegistryConfig{
MaxTotalUnreadyPercentage: 10,
OkTotalUnreadyCount: 1,
}, fakeLogRecorder, NewBackoff())
clusterState.RegisterProviders(clusterstate.NewMockMaxNodeProvisionTimeProvider(context.AutoscalingOptions.NodeGroupDefaults.MaxNodeProvisionTime))
}, fakeLogRecorder, NewBackoff(), nodegroupconfig.NewDefaultNodeGroupConfigProcessor(context.AutoscalingOptions.NodeGroupDefaults))
err := clusterState.UpdateNodes([]*apiv1.Node{regNode}, nil, now.Add(-time.Hour))
assert.NoError(t, err)