Add metrics support for imageprefetch and nodeimageset controllers

[zeroalphat]

No description provided.

[Copilot]

Pull Request Overview

This PR adds Prometheus metrics support to both the imageprefetch and nodeimageset controllers, enabling monitoring of image prefetching operations across the cluster.

Key changes include:

• New metrics package exposing gauges for image availability, pull success/failure counts, image metadata, size, and prefetch duration
• Enhanced image pull tracking to capture timing and size information
• Integration of metrics collection in both controller reconciliation loops

Reviewed Changes

Copilot reviewed 11 out of 11 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
internal/metrics/metrics.go	Defines Prometheus gauge vectors for tracking image prefetch operations and a registration function
internal/imgmanager/image_puller.go	Adds timing and size tracking to ImagePullStatus and exposes getter methods for metrics
internal/imgmanager/fake_containerd.go	Updates mock to return image size from PullImage
internal/imgmanager/containerd.go	Modifies PullImage to return image size alongside error
internal/controller/nodeimageset_controller.go	Integrates metrics recording during reconciliation and cleanup on deletion
internal/controller/imageprefetch_controller.go	Adds metrics setters/removers and calls them during reconciliation and finalization
go.mod	Moves prometheus/client_golang from indirect to direct dependency
config/manager/nodeimageset-controller.yaml	Adds metrics server configuration flags
config/manager/imageprefetch-controller.yaml	Adds metrics server configuration flags
cmd/nodeimageset-controller/main.go	Registers metrics with controller-runtime metrics registry
cmd/imageprefetch-controller/main.go	Registers metrics with controller-runtime metrics registry

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[masa213f]

Deleting metrics should be done before removing the finalizer.
If an error occurs during the update, metrics may leak depending on the error.

[zeroalphat]

Thank you.
I fixed the code to ensure that deleting metrics is done before removing the finalizer.

[masa213f]

It's better to make the label's key and value like kube-state-metrics.
The key is the resource kind, and the label is the resource name.

So then, label_replace becomes unnecessary when combined with other metrics in MetricsQL.

Suggested change

	ReadyVec = prometheus.NewGaugeVec(prometheus.GaugeOpts{
	Namespace: metricsNamespace,
	Name: "imageprefetch_ready",
	Help: "1 if the ImagePrefetch resource is ready, 0 otherwise",
	}, []string{"namespace", "name"})
	ImagePulledNodesVec = prometheus.NewGaugeVec(prometheus.GaugeOpts{
	Namespace: metricsNamespace,
	Name: "imageprefetch_image_pull_successful_nodes",
	Help: "Number of nodes where images have been successfully prefetched for thins Imageprefetch",
	}, []string{"namespace", "name"})
	ImagePullFailedNodesVec = prometheus.NewGaugeVec(prometheus.GaugeOpts{
	Namespace: metricsNamespace,
	Name: "imageprefetch_image_pull_failed_nodes",
	Help: "Number of nodes where image prefetching has failed for this ImagePrefetch",
	}, []string{"namespace", "name"})
	ReadyVec = prometheus.NewGaugeVec(prometheus.GaugeOpts{
	Namespace: metricsNamespace,
	Name: "imageprefetch_ready",
	Help: "1 if the ImagePrefetch resource is ready, 0 otherwise",
	}, []string{"namespace", "imageprefetch"})
	ImagePulledNodesVec = prometheus.NewGaugeVec(prometheus.GaugeOpts{
	Namespace: metricsNamespace,
	Name: "imageprefetch_image_pull_successful_nodes",
	Help: "Number of nodes where images have been successfully prefetched for thins Imageprefetch",
	}, []string{"namespace", "imageprefetch"})
	ImagePullFailedNodesVec = prometheus.NewGaugeVec(prometheus.GaugeOpts{
	Namespace: metricsNamespace,
	Name: "imageprefetch_image_pull_failed_nodes",
	Help: "Number of nodes where image prefetching has failed for this ImagePrefetch",
	}, []string{"namespace", "imageprefetch"})

[zeroalphat]

I fixed it to change the metrics labels to be like kube-state-metrics.

[masa213f]

ditto.
It is better to use similar label keys to kube-state-metrics.

Suggested change

	ImageInfoVec = prometheus.NewGaugeVec(prometheus.GaugeOpts{
	Namespace: metricsNamespace,
	Name: "nodeimageset_image_info",
	Help: "Information about NodeImageSet image",
	}, []string{"name", "image_name", "registry_policy", "node_name"})
	ImageSizeBytesVec = prometheus.NewGaugeVec(prometheus.GaugeOpts{
	Namespace: metricsNamespace,
	Name: "nodeimageset_image_size_bytes",
	Help: "Size of images in NodeImageSets in bytes",
	}, []string{"name", "image_name", "node_name"})
	ImagePrefetchDurationSecondsVec = prometheus.NewGaugeVec(prometheus.GaugeOpts{
	Namespace: metricsNamespace,
	Name: "nodeimageset_image_prefetch_duration_seconds",
	Help: "Duration taken to prefetch images in NodeImageSets in seconds",
	}, []string{"name", "image_name", "node_name"})
	ImageInfoVec = prometheus.NewGaugeVec(prometheus.GaugeOpts{
	Namespace: metricsNamespace,
	Name: "nodeimageset_image_info",
	Help: "Information about NodeImageSet image",
	}, []string{"nodeimageset", "image", "registry_policy", "node"})
	ImageSizeBytesVec = prometheus.NewGaugeVec(prometheus.GaugeOpts{
	Namespace: metricsNamespace,
	Name: "nodeimageset_image_size_bytes",
	Help: "Size of images in NodeImageSets in bytes",
	}, []string{"nodeimageset", "image", "node"})
	ImagePrefetchDurationSecondsVec = prometheus.NewGaugeVec(prometheus.GaugeOpts{
	Namespace: metricsNamespace,
	Name: "nodeimageset_image_prefetch_duration_seconds",
	Help: "Duration taken to prefetch images in NodeImageSets in seconds",
	}, []string{"nodeimageset", "image", "node"})

[masa213f]

LGTM