Skip to content

Add performance benchmark for the CA RunOnce control loop #9237

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
Choraden wants to merge 9 commits into
base: master
Choose a base branch
from
+540 −14
Open

Add performance benchmark for the CA RunOnce control loop #9237

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
95787cd
Generalize MustCreateManager to support testing.TB
Choraden Feb 11, 2026
5f74be6
Introduce performance benchmarking for the RunOnce control loop
Choraden Feb 4, 2026
6bd19d7
Add CPU profiling support to RunOnce benchmarks
Choraden Feb 11, 2026
75f49a1
Disable predicate parallelism in benchmarks for stability and reprodu…
Choraden Feb 13, 2026
dec0f3f
Stabilize benchmark results by disabling GC
Choraden Feb 13, 2026
2c80c57
Silence benchmark logs to avoid output noise
Choraden Feb 13, 2026
d7eac11
Add no-op event recorder to benchmarks to avoid dropping events
Choraden Feb 13, 2026
f6e089e
Use fast scale-up in RunOnce benchmarks to reduce CPU noise
Choraden Feb 13, 2026
2c85ed6
Use fast tainting reactors in RunOnce benchmark
Choraden Feb 13, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
33 changes: 22 additions & 11 deletions cluster-autoscaler/builder/autoscaler.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -27,6 +27,7 @@ import (
"k8s.io/autoscaler/cluster-autoscaler/capacitybuffer/fakepods"
"k8s.io/autoscaler/cluster-autoscaler/cloudprovider"
"k8s.io/autoscaler/cluster-autoscaler/config"
cacontext "k8s.io/autoscaler/cluster-autoscaler/context"
"k8s.io/autoscaler/cluster-autoscaler/core"
coreoptions "k8s.io/autoscaler/cluster-autoscaler/core/options"
"k8s.io/autoscaler/cluster-autoscaler/core/podlistprocessor"
Expand Down Expand Up @@ -75,6 +76,7 @@ type AutoscalerBuilder struct {
podObserver *loop.UnschedulablePodObserver
cloudProvider cloudprovider.CloudProvider
informerFactory informers.SharedInformerFactory
kubeClients *cacontext.AutoscalingKubeClients
}

// New creates a builder with default options.
Expand Down Expand Up @@ -120,6 +122,14 @@ func (b *AutoscalerBuilder) WithInformerFactory(f informers.SharedInformerFactor
return b
}

// WithAutoscalingKubeClients allows injecting autoscaling kube clients.
// It is not needed for most use-cases.
// Once used, it has to be in sync with the object provided in WithKubeClient and WithInformerFactory.
func (b *AutoscalerBuilder) WithAutoscalingKubeClients(kubeClients *cacontext.AutoscalingKubeClients) *AutoscalerBuilder {
Copy link
Collaborator

@towca towca Mar 10, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

AutoscalingKubeClients is normally built based on the KubeClient and InformerFactory, both of which are already injectable (and as far I can see they need to be injected). Why do we need to make AutoscalingKubeClients directly injectable too? It seems like we'd always want to have the AutoscalingKubeClients field in sync with the other two. Otherwise we'd get different behavior for components that use the client/informers directly than for components that use AutoscalingKubeClients.

Copy link
Collaborator

@towca towca Mar 10, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ah okay, after reading some more I see why - we want to change the log recorder. Could you add a warning to the method comment that this is not needed for most use-cases, and if you're using it it has to be in sync with the objects provided in WithKubeClient() and WithInformerFactory()?

Copy link
Contributor Author

@Choraden Choraden Mar 13, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Added a comment to the method.

b.kubeClients = kubeClients
return b
}

// Build constructs the Autoscaler based on the provided configuration.
func (b *AutoscalerBuilder) Build(ctx context.Context) (core.Autoscaler, *loop.LoopTrigger, error) {
// Get AutoscalingOptions from flags.
Expand Down Expand Up @@ -147,17 +157,18 @@ func (b *AutoscalerBuilder) Build(ctx context.Context) (core.Autoscaler, *loop.L

var snapshotStore clustersnapshot.ClusterSnapshotStore = store.NewDeltaSnapshotStore(autoscalingOptions.ClusterSnapshotParallelism)
opts := coreoptions.AutoscalerOptions{
AutoscalingOptions: autoscalingOptions,
FrameworkHandle: fwHandle,
ClusterSnapshot: predicate.NewPredicateSnapshot(snapshotStore, fwHandle, autoscalingOptions.DynamicResourceAllocationEnabled, autoscalingOptions.PredicateParallelism, autoscalingOptions.CSINodeAwareSchedulingEnabled),
KubeClient: b.kubeClient,
InformerFactory: b.informerFactory,
DebuggingSnapshotter: b.debuggingSnapshotter,
DeleteOptions: deleteOptions,
DrainabilityRules: drainabilityRules,
ScaleUpOrchestrator: orchestrator.New(),
KubeClientNew: b.manager.GetClient(),
KubeCache: b.manager.GetCache(),
AutoscalingOptions: autoscalingOptions,
FrameworkHandle: fwHandle,
ClusterSnapshot: predicate.NewPredicateSnapshot(snapshotStore, fwHandle, autoscalingOptions.DynamicResourceAllocationEnabled, autoscalingOptions.PredicateParallelism, autoscalingOptions.CSINodeAwareSchedulingEnabled),
KubeClient: b.kubeClient,
InformerFactory: b.informerFactory,
AutoscalingKubeClients: b.kubeClients,
DebuggingSnapshotter: b.debuggingSnapshotter,
DeleteOptions: deleteOptions,
DrainabilityRules: drainabilityRules,
ScaleUpOrchestrator: orchestrator.New(),
KubeClientNew: b.manager.GetClient(),
KubeCache: b.manager.GetCache(),
}

opts.Processors = ca_processors.DefaultProcessors(autoscalingOptions)
Expand Down
19 changes: 18 additions & 1 deletion cluster-autoscaler/cloudprovider/test/fake_cloud_provider.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,14 +18,15 @@ package test

import (
"fmt"
"sync"

apiv1 "k8s.io/api/core/v1"
"k8s.io/apimachinery/pkg/api/resource"
"k8s.io/autoscaler/cluster-autoscaler/cloudprovider"
"k8s.io/autoscaler/cluster-autoscaler/config"
"k8s.io/autoscaler/cluster-autoscaler/simulator/framework"
"k8s.io/autoscaler/cluster-autoscaler/utils/errors"
fakek8s "k8s.io/autoscaler/cluster-autoscaler/utils/fake"
"sync"
)

const (
Expand Down Expand Up @@ -153,6 +154,22 @@ func WithNode(node *apiv1.Node) NodeGroupOption {
}
}

// WithNGSize sets the minimum and maximum size of the node group.
func WithNGSize(min, max int) NodeGroupOption {
return func(n *NodeGroup) {
n.minSize = min
n.maxSize = max

}
}

// WithTemplate sets the node template for the node group.
func WithTemplate(template *framework.NodeInfo) NodeGroupOption {
return func(n *NodeGroup) {
n.template = template
}
}

// AddNodeGroup is a helper for tests to add a group with its template.
func (c *CloudProvider) AddNodeGroup(id string, opts ...NodeGroupOption) {
c.Lock()
Expand Down
Loading
Loading