investigate, refactor preload tar balls

[medyagh]

when doing preload PR for containerd I had a couple ideas and question

if user adds images using cache add

if they add one couple images using cache add ( as part of a development workflow)

if they start agian are we invoking a reset on all pull ?

// DockerImagesPreloaded returns true if all images have been preloaded
func DockerImagesPreloaded(runner command.Runner, images []string) bool {
	rr, err := runner.RunCmd(exec.Command("docker", "images", "--format", "{{.Repository}}:{{.Tag}}"))
	if err != nil {
		return false
	}
	preloadedImages := map[string]struct{}{}
	for _, i := range strings.Split(rr.Stdout.String(), "\n") {
		preloadedImages[i] = struct{}{}
	}

	glog.Infof("Got preloaded images: %s", rr.Output())

	// Make sure images == imgs
	for _, i := range images {
		if _, ok := preloadedImages[i]; !ok {
			glog.Infof("%s wasn't preloaded", i)
			return false
		}
	}
	return true
}

can we refactor to DockerImagesPreloaded to return list of not preloaded images ( instead of bool)

so we only pull and load the images that do not exist as opposed load all images again ?

[medyagh]

alternatively we could separate the tarballs for kubernetes core images and aux images and binaries

[fejta-bot]

Issues go stale after 90d of inactivity.
Mark the issue as fresh with /remove-lifecycle stale.
Stale issues rot after an additional 30d of inactivity and eventually close.

If this issue is safe to close now please do so with /close.

Send feedback to sig-testing, kubernetes/test-infra and/or fejta.
/lifecycle stale

[fejta-bot]

Stale issues rot after 30d of inactivity.
Mark the issue as fresh with /remove-lifecycle rotten.
Rotten issues close after an additional 30d of inactivity.

If this issue is safe to close now please do so with /close.

Send feedback to sig-testing, kubernetes/test-infra and/or fejta.
/lifecycle rotten

[priyawadhwa]

This would improve performance, but before we do LoadImage we check if the image already exists in the daemon in the needsTransfer function --

minikube/pkg/minikube/machine/cache_images.go

Line 139 in 070cbb9

func needsTransfer(imgClient *client.Client, imgName string, cr cruntime.Manager) error {

My guess is we're looking at a small performance improvement (<1s), but it should be easy enough to do, especially if we compare preloaded images to images that need to be loaded here:

minikube/pkg/minikube/machine/cache_images.go

Line 73 in 070cbb9

if cr.ImagesPreloaded(images) {

[priyawadhwa]

If we end up implementing the idea mentioned here -- #10438 (comment) -- then this issue is obsolete.