Kubelet: Cache image history to eliminate the performance regression

pkg/kubelet/dockertools/fake_docker_client.go

@@ -30,7 +30,6 @@ import (
 	dockercontainer "github.com/docker/engine-api/types/container"
 
 	"k8s.io/kubernetes/pkg/api"
-	"k8s.io/kubernetes/pkg/util/sets"
 )
 
 // FakeDockerClient is a simple fake docker client, so that kubelet can be run for testing without requiring a real docker setup.
@@ -49,7 +48,6 @@ type FakeDockerClient struct {
 	Created         []string
 	Stopped         []string
 	Removed         []string
-	RemovedImages   sets.String
 	VersionInfo     dockertypes.Version
 	Information     dockertypes.Info
 	ExecInspect     *dockertypes.ContainerExecInspect
@@ -68,10 +66,9 @@ func NewFakeDockerClient() *FakeDockerClient {
 
 func NewFakeDockerClientWithVersion(version, apiVersion string) *FakeDockerClient {
 	return &FakeDockerClient{
-		VersionInfo:   dockertypes.Version{Version: version, APIVersion: apiVersion},
-		Errors:        make(map[string]error),
-		RemovedImages: sets.String{},
-		ContainerMap:  make(map[string]*dockertypes.ContainerJSON),
+		VersionInfo:  dockertypes.Version{Version: version, APIVersion: apiVersion},
+		Errors:       make(map[string]error),
+		ContainerMap: make(map[string]*dockertypes.ContainerJSON),
 	}
 }
 
@@ -471,14 +468,20 @@ func (f *FakeDockerClient) InspectExec(id string) (*dockertypes.ContainerExecIns
 }
 
 func (f *FakeDockerClient) ListImages(opts dockertypes.ImageListOptions) ([]dockertypes.Image, error) {
+	f.called = append(f.called, "list_images")
 	err := f.popError("list_images")
 	return f.Images, err
 }
 
 func (f *FakeDockerClient) RemoveImage(image string, opts dockertypes.ImageRemoveOptions) ([]dockertypes.ImageDelete, error) {
 	err := f.popError("remove_image")
 	if err == nil {
-		f.RemovedImages.Insert(image)
+		for i := range f.Images {
+			if f.Images[i].ID == image {
+				f.Images = append(f.Images[:i], f.Images[i+1:]...)
+				break
+			}
+		}
 	}
 	return []dockertypes.ImageDelete{{Deleted: image}}, err
 }
@@ -538,6 +541,7 @@ func (f *FakeDockerPuller) IsImagePresent(name string) (bool, error) {
 func (f *FakeDockerClient) ImageHistory(id string) ([]dockertypes.ImageHistory, error) {
 	f.Lock()
 	defer f.Unlock()
+	f.called = append(f.called, "image_history")
 	history := f.ImageHistoryMap[id]
 	return history, nil
 }

pkg/kubelet/dockertools/images.go

@@ -18,54 +18,85 @@ package dockertools
 
 import (
 	"fmt"
+	"sync"
 
 	"github.com/golang/glog"
 
 	dockertypes "github.com/docker/engine-api/types"
 	runtime "k8s.io/kubernetes/pkg/kubelet/container"
-	"k8s.io/kubernetes/pkg/util/sets"
 )
 
 // imageStatsProvider exposes stats about all images currently available.
 type imageStatsProvider struct {
+	sync.Mutex
+	// layers caches the current layers, key is the layer ID.
+	layers map[string]*dockertypes.ImageHistory
+	// imageToLayerIDs maps image to its layer IDs.
+	imageToLayerIDs map[string][]string
 	// Docker remote API client
 	c DockerInterface
 }
 
+func newImageStatsProvider(c DockerInterface) *imageStatsProvider {
+	return &imageStatsProvider{
+		layers:          make(map[string]*dockertypes.ImageHistory),
+		imageToLayerIDs: make(map[string][]string),
+		c:               c,
+	}
+}
+
 func (isp *imageStatsProvider) ImageStats() (*runtime.ImageStats, error) {
 	images, err := isp.c.ListImages(dockertypes.ImageListOptions{})
 	if err != nil {
 		return nil, fmt.Errorf("failed to list docker images - %v", err)
 	}
-	// A map of all the image layers to its corresponding size.
-	imageMap := sets.NewString()
-	ret := &runtime.ImageStats{}
+	// Take the lock to protect the cache
+	isp.Lock()
+	defer isp.Unlock()
+	// Create new cache each time, this is a little more memory consuming, but:
+	// * ImageStats is only called every 10 seconds
+	// * We use pointers and reference to copy cache elements.
+	// The memory usage should be acceptable.
+	// TODO(random-liu): Add more logic to implement in place cache update.
+	newLayers := make(map[string]*dockertypes.ImageHistory)
+	newImageToLayerIDs := make(map[string][]string)
 	for _, image := range images {
-		// Get information about the various layers of each docker image.
-		history, err := isp.c.ImageHistory(image.ID)
-		if err != nil {
-			glog.V(2).Infof("failed to get history of docker image %v - %v", image, err)
-			continue
-		}
-		// Store size information of each layer.
-		for _, layer := range history {
-			// Skip empty layers.
-			if layer.Size == 0 {
-				glog.V(10).Infof("skipping image layer %v with size 0", layer)
+		layerIDs, ok := isp.imageToLayerIDs[image.ID]
+		if !ok {
+			// Get information about the various layers of the given docker image.
+			history, err := isp.c.ImageHistory(image.ID)
+			if err != nil {
+				// Skip the image and inspect again in next ImageStats if the image is still there
+				glog.V(2).Infof("failed to get history of docker image %+v - %v", image, err)
 				continue
 			}
-			key := layer.ID
-			// Some of the layers are empty.
-			// We are hoping that these layers are unique to each image.
-			// Still keying with the CreatedBy field to be safe.
-			if key == "" || key == "<missing>" {
-				key = key + layer.CreatedBy
+			// Cache each layer
+			for i := range history {
+				layer := &history[i]
+				key := layer.ID
+				// Some of the layers are empty.
+				// We are hoping that these layers are unique to each image.
+				// Still keying with the CreatedBy field to be safe.
+				if key == "" || key == "<missing>" {
+					key = key + layer.CreatedBy
+				}
+				layerIDs = append(layerIDs, key)
+				newLayers[key] = layer
 			}
-			if !imageMap.Has(key) {
-				ret.TotalStorageBytes += uint64(layer.Size)
+		} else {
+			for _, layerID := range layerIDs {
+				newLayers[layerID] = isp.layers[layerID]
 			}
-			imageMap.Insert(key)
 		}
+		newImageToLayerIDs[image.ID] = layerIDs
+	}
+	ret := &runtime.ImageStats{}
+	// Calculate the total storage bytes
+	for _, layer := range newLayers {
+		ret.TotalStorageBytes += uint64(layer.Size)
 	}
+	// Update current cache
+	isp.layers = newLayers
+	isp.imageToLayerIDs = newImageToLayerIDs
 	return ret, nil
 }