diff options
author | OpenShift Merge Robot <openshift-merge-robot@users.noreply.github.com> | 2019-04-25 11:16:45 -0700 |
---|---|---|
committer | GitHub <noreply@github.com> | 2019-04-25 11:16:45 -0700 |
commit | e900e95f9aa6a22eeef9f78573dfb88df2f10d49 (patch) | |
tree | ea8137507f196ad7bdd3a080e4b4e769ce6ecae7 | |
parent | 36f479097d6617bd4a078144a8709c19f865eb8f (diff) | |
parent | 4daac4736e3476211de7c1672bb5161191666a91 (diff) | |
download | podman-e900e95f9aa6a22eeef9f78573dfb88df2f10d49.tar.gz podman-e900e95f9aa6a22eeef9f78573dfb88df2f10d49.tar.bz2 podman-e900e95f9aa6a22eeef9f78573dfb88df2f10d49.zip |
Merge pull request #2990 from nalind/image-parent-child
image: rework parent/child/history matching
-rw-r--r-- | cmd/podman/images.go | 2 | ||||
-rw-r--r-- | cmd/podman/images_prune.go | 2 | ||||
-rw-r--r-- | cmd/podman/rmi.go | 2 | ||||
-rw-r--r-- | cmd/podman/system_df.go | 10 | ||||
-rw-r--r-- | cmd/podman/system_prune.go | 2 | ||||
-rw-r--r-- | libpod/image/image.go | 243 | ||||
-rw-r--r-- | libpod/image/image_test.go | 2 | ||||
-rw-r--r-- | libpod/image/prune.go | 6 | ||||
-rw-r--r-- | libpod/runtime_img.go | 6 | ||||
-rw-r--r-- | pkg/adapter/runtime.go | 4 | ||||
-rw-r--r-- | pkg/adapter/runtime_remote.go | 4 | ||||
-rw-r--r-- | pkg/varlinkapi/images.go | 7 |
12 files changed, 228 insertions, 62 deletions
diff --git a/cmd/podman/images.go b/cmd/podman/images.go index f584c1131..41aa213a8 100644 --- a/cmd/podman/images.go +++ b/cmd/podman/images.go @@ -243,7 +243,7 @@ func getImagesTemplateOutput(ctx context.Context, images []*adapter.ContainerIma // If all is false and the image doesn't have a name, check to see if the top layer of the image is a parent // to another image's top layer. If it is, then it is an intermediate image so don't print out if the --all flag // is not set. - isParent, err := img.IsParent() + isParent, err := img.IsParent(ctx) if err != nil { logrus.Errorf("error checking if image is a parent %q: %v", img.ID(), err) } diff --git a/cmd/podman/images_prune.go b/cmd/podman/images_prune.go index 84181d0a2..c522c8b15 100644 --- a/cmd/podman/images_prune.go +++ b/cmd/podman/images_prune.go @@ -45,7 +45,7 @@ func pruneImagesCmd(c *cliconfig.PruneImagesValues) error { // Call prune; if any cids are returned, print them and then // return err in case an error also came up - pruneCids, err := runtime.PruneImages(c.All) + pruneCids, err := runtime.PruneImages(getContext(), c.All) if len(pruneCids) > 0 { for _, cid := range pruneCids { fmt.Println(cid) diff --git a/cmd/podman/rmi.go b/cmd/podman/rmi.go index 7ec875d5b..be7c81dab 100644 --- a/cmd/podman/rmi.go +++ b/cmd/podman/rmi.go @@ -97,7 +97,7 @@ func rmiCmd(c *cliconfig.RmiValues) error { return errors.New("unable to delete all images; re-run the rmi command again.") } for _, i := range imagesToDelete { - isParent, err := i.IsParent() + isParent, err := i.IsParent(ctx) if err != nil { return err } diff --git a/cmd/podman/system_df.go b/cmd/podman/system_df.go index 16a8ad120..aa0ead022 100644 --- a/cmd/podman/system_df.go +++ b/cmd/podman/system_df.go @@ -201,7 +201,7 @@ func imageUniqueSize(ctx context.Context, images []*image.Image) (map[string]uin for _, img := range images { parentImg := img for { - next, err := parentImg.GetParent() + next, err := parentImg.GetParent(ctx) if err != nil { return nil, errors.Wrapf(err, "error getting parent of image %s", parentImg.ID()) } @@ -246,11 +246,11 @@ func getImageDiskUsage(ctx context.Context, images []*image.Image, imageUsedbyCi unreclaimableSize += imageUsedSize(img, imgUniqueSizeMap, imageUsedbyCintainerMap, imageUsedbyActiveContainerMap) - isParent, err := img.IsParent() + isParent, err := img.IsParent(ctx) if err != nil { return imageDiskUsage, err } - parent, err := img.GetParent() + parent, err := img.GetParent(ctx) if err != nil { return imageDiskUsage, errors.Wrapf(err, "error getting parent of image %s", img.ID()) } @@ -437,11 +437,11 @@ func getImageVerboseDiskUsage(ctx context.Context, images []*image.Image, images return imagesVerboseDiskUsage, errors.Wrapf(err, "error getting unique size of images") } for _, img := range images { - isParent, err := img.IsParent() + isParent, err := img.IsParent(ctx) if err != nil { return imagesVerboseDiskUsage, errors.Wrapf(err, "error checking if %s is a parent images", img.ID()) } - parent, err := img.GetParent() + parent, err := img.GetParent(ctx) if err != nil { return imagesVerboseDiskUsage, errors.Wrapf(err, "error getting parent of image %s", img.ID()) } diff --git a/cmd/podman/system_prune.go b/cmd/podman/system_prune.go index 8900e2644..2c1c5607a 100644 --- a/cmd/podman/system_prune.go +++ b/cmd/podman/system_prune.go @@ -110,7 +110,7 @@ Are you sure you want to continue? [y/N] `, volumeString) // Call prune; if any cids are returned, print them and then // return err in case an error also came up - pruneCids, err := runtime.PruneImages(c.All) + pruneCids, err := runtime.PruneImages(ctx, c.All) if len(pruneCids) > 0 { fmt.Println("Deleted Images") for _, cid := range pruneCids { diff --git a/libpod/image/image.go b/libpod/image/image.go index 757d034a2..0b0002d42 100644 --- a/libpod/image/image.go +++ b/libpod/image/image.go @@ -353,8 +353,8 @@ func (i *Image) TopLayer() string { // outside the context of images // TODO: the force param does nothing as of now. Need to move container // handling logic here eventually. -func (i *Image) Remove(force bool) error { - parent, err := i.GetParent() +func (i *Image) Remove(ctx context.Context, force bool) error { + parent, err := i.GetParent(ctx) if err != nil { return err } @@ -363,11 +363,11 @@ func (i *Image) Remove(force bool) error { } i.newImageEvent(events.Remove) for parent != nil { - nextParent, err := parent.GetParent() + nextParent, err := parent.GetParent(ctx) if err != nil { return err } - children, err := parent.GetChildren() + children, err := parent.GetChildren(ctx) if err != nil { return err } @@ -679,7 +679,8 @@ type History struct { Comment string `json:"comment"` } -// History gets the history of an image and information about its layers +// History gets the history of an image and the IDs of images that are part of +// its history func (i *Image) History(ctx context.Context) ([]*History, error) { img, err := i.toImageRef(ctx) if err != nil { @@ -690,31 +691,92 @@ func (i *Image) History(ctx context.Context) ([]*History, error) { return nil, err } - // Get the IDs of the images making up the history layers - // if the images exist locally in the store + // Use our layers list to find images that use one of them as its + // topmost layer. + interestingLayers := make(map[string]bool) + layer, err := i.imageruntime.store.Layer(i.TopLayer()) + if err != nil { + return nil, err + } + for layer != nil { + interestingLayers[layer.ID] = true + if layer.Parent == "" { + break + } + layer, err = i.imageruntime.store.Layer(layer.Parent) + if err != nil { + return nil, err + } + } + + // Get the IDs of the images that share some of our layers. Hopefully + // this step means that we'll be able to avoid reading the + // configuration of every single image in local storage later on. images, err := i.imageruntime.GetImages() if err != nil { return nil, errors.Wrapf(err, "error getting images from store") } - imageIDs := []string{i.ID()} - if err := i.historyLayerIDs(i.TopLayer(), images, &imageIDs); err != nil { - return nil, errors.Wrap(err, "error getting image IDs for layers in history") + interestingImages := make([]*Image, 0, len(images)) + for i := range images { + if interestingLayers[images[i].TopLayer()] { + interestingImages = append(interestingImages, images[i]) + } + } + + // Build a list of image IDs that correspond to our history entries. + historyImages := make([]*Image, len(oci.History)) + if len(oci.History) > 0 { + // The starting image shares its whole history with itself. + historyImages[len(historyImages)-1] = i + for i := range interestingImages { + image, err := images[i].ociv1Image(ctx) + if err != nil { + return nil, errors.Wrapf(err, "error getting image configuration for image %q", images[i].ID()) + } + // If the candidate has a longer history or no history + // at all, then it doesn't share the portion of our + // history that we're interested in matching with other + // images. + if len(image.History) == 0 || len(image.History) > len(historyImages) { + continue + } + // If we don't include all of the layers that the + // candidate image does (i.e., our rootfs didn't look + // like its rootfs at any point), then it can't be part + // of our history. + if len(image.RootFS.DiffIDs) > len(oci.RootFS.DiffIDs) { + continue + } + candidateLayersAreUsed := true + for i := range image.RootFS.DiffIDs { + if image.RootFS.DiffIDs[i] != oci.RootFS.DiffIDs[i] { + candidateLayersAreUsed = false + break + } + } + if !candidateLayersAreUsed { + continue + } + // If the candidate's entire history is an initial + // portion of our history, then we're based on it, + // either directly or indirectly. + sharedHistory := historiesMatch(oci.History, image.History) + if sharedHistory == len(image.History) { + historyImages[sharedHistory-1] = images[i] + } + } } var ( - imageID string - imgIDCount = 0 size int64 sizeCount = 1 allHistory []*History ) for i := len(oci.History) - 1; i >= 0; i-- { - if imgIDCount < len(imageIDs) { - imageID = imageIDs[imgIDCount] - imgIDCount++ - } else { - imageID = "<missing>" + imageID := "<missing>" + if historyImages[i] != nil { + imageID = historyImages[i].ID() } if !oci.History[i].EmptyLayer { size = img.LayerInfos()[len(img.LayerInfos())-sizeCount].Size @@ -1006,26 +1068,110 @@ func splitString(input string) string { // IsParent goes through the layers in the store and checks if i.TopLayer is // the parent of any other layer in store. Double check that image with that // layer exists as well. -func (i *Image) IsParent() (bool, error) { - children, err := i.GetChildren() +func (i *Image) IsParent(ctx context.Context) (bool, error) { + children, err := i.getChildren(ctx, 1) if err != nil { return false, err } return len(children) > 0, nil } +// historiesMatch returns the number of entries in the histories which have the +// same contents +func historiesMatch(a, b []imgspecv1.History) int { + i := 0 + for i < len(a) && i < len(b) { + if a[i].Created != nil && b[i].Created == nil { + return i + } + if a[i].Created == nil && b[i].Created != nil { + return i + } + if a[i].Created != nil && b[i].Created != nil { + if !a[i].Created.Equal(*(b[i].Created)) { + return i + } + } + if a[i].CreatedBy != b[i].CreatedBy { + return i + } + if a[i].Author != b[i].Author { + return i + } + if a[i].Comment != b[i].Comment { + return i + } + if a[i].EmptyLayer != b[i].EmptyLayer { + return i + } + i++ + } + return i +} + +// areParentAndChild checks diff ID and history in the two images and return +// true if the second should be considered to be directly based on the first +func areParentAndChild(parent, child *imgspecv1.Image) bool { + // the child and candidate parent should share all of the + // candidate parent's diff IDs, which together would have + // controlled which layers were used + if len(parent.RootFS.DiffIDs) > len(child.RootFS.DiffIDs) { + return false + } + childUsesCandidateDiffs := true + for i := range parent.RootFS.DiffIDs { + if child.RootFS.DiffIDs[i] != parent.RootFS.DiffIDs[i] { + childUsesCandidateDiffs = false + break + } + } + if !childUsesCandidateDiffs { + return false + } + // the child should have the same history as the parent, plus + // one more entry + if len(parent.History)+1 != len(child.History) { + return false + } + if historiesMatch(parent.History, child.History) != len(parent.History) { + return false + } + return true +} + // GetParent returns the image ID of the parent. Return nil if a parent is not found. -func (i *Image) GetParent() (*Image, error) { +func (i *Image) GetParent(ctx context.Context) (*Image, error) { images, err := i.imageruntime.GetImages() if err != nil { return nil, err } - layer, err := i.imageruntime.store.Layer(i.TopLayer()) + childLayer, err := i.imageruntime.store.Layer(i.TopLayer()) + if err != nil { + return nil, err + } + // fetch the configuration for the child image + child, err := i.ociv1Image(ctx) if err != nil { return nil, err } for _, img := range images { - if img.TopLayer() == layer.Parent { + if img.ID() == i.ID() { + continue + } + candidateLayer := img.TopLayer() + // as a child, our top layer is either the candidate parent's + // layer, or one that's derived from it, so skip over any + // candidate image where we know that isn't the case + if candidateLayer != childLayer.Parent && candidateLayer != childLayer.ID { + continue + } + // fetch the configuration for the candidate image + candidate, err := img.ociv1Image(ctx) + if err != nil { + return nil, err + } + // compare them + if areParentAndChild(candidate, child) { return img, nil } } @@ -1033,36 +1179,53 @@ func (i *Image) GetParent() (*Image, error) { } // GetChildren returns a list of the imageIDs that depend on the image -func (i *Image) GetChildren() ([]string, error) { +func (i *Image) GetChildren(ctx context.Context) ([]string, error) { + return i.getChildren(ctx, 0) +} + +// getChildren returns a list of at most "max" imageIDs that depend on the image +func (i *Image) getChildren(ctx context.Context, max int) ([]string, error) { var children []string images, err := i.imageruntime.GetImages() if err != nil { return nil, err } - layers, err := i.imageruntime.store.Layers() + + // fetch the configuration for the parent image + parent, err := i.ociv1Image(ctx) if err != nil { return nil, err } + parentLayer := i.TopLayer() - for _, layer := range layers { - if layer.Parent == i.TopLayer() { - if imageID := getImageOfTopLayer(images, layer.ID); len(imageID) > 0 { - children = append(children, imageID...) - } - } - } - return children, nil -} - -// getImageOfTopLayer returns the image ID where layer is the top layer of the image -func getImageOfTopLayer(images []*Image, layer string) []string { - var matches []string for _, img := range images { - if img.TopLayer() == layer { - matches = append(matches, img.ID()) + if img.ID() == i.ID() { + continue + } + candidateLayer, err := img.Layer() + if err != nil { + return nil, err + } + // if this image's top layer is not our top layer, and is not + // based on our top layer, we can skip it + if candidateLayer.Parent != parentLayer && candidateLayer.ID != parentLayer { + continue + } + // fetch the configuration for the candidate image + candidate, err := img.ociv1Image(ctx) + if err != nil { + return nil, err + } + // compare them + if areParentAndChild(parent, candidate) { + children = append(children, img.ID()) + } + // if we're not building an exhaustive list, maybe we're done? + if max > 0 && len(children) >= max { + break } } - return matches + return children, nil } // InputIsID returns a bool if the user input for an image diff --git a/libpod/image/image_test.go b/libpod/image/image_test.go index 075ba119d..d917e3205 100644 --- a/libpod/image/image_test.go +++ b/libpod/image/image_test.go @@ -139,7 +139,7 @@ func TestImage_New(t *testing.T) { newImage, err := ir.New(context.Background(), img, "", "", writer, nil, SigningOptions{}, false, nil) assert.NoError(t, err) assert.NotEqual(t, newImage.ID(), "") - err = newImage.Remove(false) + err = newImage.Remove(context.Background(), false) assert.NoError(t, err) } diff --git a/libpod/image/prune.go b/libpod/image/prune.go index 5bd3c2c99..a4f8a0c9f 100644 --- a/libpod/image/prune.go +++ b/libpod/image/prune.go @@ -1,6 +1,8 @@ package image import ( + "context" + "github.com/containers/libpod/libpod/events" "github.com/pkg/errors" ) @@ -34,14 +36,14 @@ func (ir *Runtime) GetPruneImages(all bool) ([]*Image, error) { // PruneImages prunes dangling and optionally all unused images from the local // image store -func (ir *Runtime) PruneImages(all bool) ([]string, error) { +func (ir *Runtime) PruneImages(ctx context.Context, all bool) ([]string, error) { var prunedCids []string pruneImages, err := ir.GetPruneImages(all) if err != nil { return nil, errors.Wrap(err, "unable to get images to prune") } for _, p := range pruneImages { - if err := p.Remove(true); err != nil { + if err := p.Remove(ctx, true); err != nil { return nil, errors.Wrap(err, "failed to prune image") } defer p.newImageEvent(events.Prune) diff --git a/libpod/runtime_img.go b/libpod/runtime_img.go index 02f925fc6..5e9f65acc 100644 --- a/libpod/runtime_img.go +++ b/libpod/runtime_img.go @@ -57,7 +57,7 @@ func (r *Runtime) RemoveImage(ctx context.Context, img *image.Image, force bool) } } - hasChildren, err := img.IsParent() + hasChildren, err := img.IsParent(ctx) if err != nil { return "", err } @@ -82,12 +82,12 @@ func (r *Runtime) RemoveImage(ctx context.Context, img *image.Image, force bool) // reponames and no force is applied, we error out. return "", fmt.Errorf("unable to delete %s (must force) - image is referred to in multiple tags", img.ID()) } - err = img.Remove(force) + err = img.Remove(ctx, force) if err != nil && errors.Cause(err) == storage.ErrImageUsedByContainer { if errStorage := r.rmStorageContainers(force, img); errStorage == nil { // Containers associated with the image should be deleted now, // let's try removing the image again. - err = img.Remove(force) + err = img.Remove(ctx, force) } else { err = errStorage } diff --git a/pkg/adapter/runtime.go b/pkg/adapter/runtime.go index 753f7c944..ca9678265 100644 --- a/pkg/adapter/runtime.go +++ b/pkg/adapter/runtime.go @@ -119,8 +119,8 @@ func (r *LocalRuntime) RemoveImage(ctx context.Context, img *ContainerImage, for } // PruneImages is wrapper into PruneImages within the image pkg -func (r *LocalRuntime) PruneImages(all bool) ([]string, error) { - return r.ImageRuntime().PruneImages(all) +func (r *LocalRuntime) PruneImages(ctx context.Context, all bool) ([]string, error) { + return r.ImageRuntime().PruneImages(ctx, all) } // Export is a wrapper to container export to a tarfile diff --git a/pkg/adapter/runtime_remote.go b/pkg/adapter/runtime_remote.go index dcb0924ce..29ee821e0 100644 --- a/pkg/adapter/runtime_remote.go +++ b/pkg/adapter/runtime_remote.go @@ -256,7 +256,7 @@ func (r *LocalRuntime) New(ctx context.Context, name, signaturePolicyPath, authf // IsParent goes through the layers in the store and checks if i.TopLayer is // the parent of any other layer in store. Double check that image with that // layer exists as well. -func (ci *ContainerImage) IsParent() (bool, error) { +func (ci *ContainerImage) IsParent(context.Context) (bool, error) { return ci.remoteImage.isParent, nil } @@ -338,7 +338,7 @@ func (ci *ContainerImage) History(ctx context.Context) ([]*image.History, error) } // PruneImages is the wrapper call for a remote-client to prune images -func (r *LocalRuntime) PruneImages(all bool) ([]string, error) { +func (r *LocalRuntime) PruneImages(ctx context.Context, all bool) ([]string, error) { return iopodman.ImagesPrune().Call(r.Conn, all) } diff --git a/pkg/varlinkapi/images.go b/pkg/varlinkapi/images.go index 470eadaeb..cecddf6b3 100644 --- a/pkg/varlinkapi/images.go +++ b/pkg/varlinkapi/images.go @@ -4,6 +4,7 @@ package varlinkapi import ( "bytes" + "context" "encoding/json" "fmt" "io" @@ -49,7 +50,7 @@ func (i *LibpodAPI) ListImages(call iopodman.VarlinkCall) error { } size, _ := image.Size(getContext()) - isParent, err := image.IsParent() + isParent, err := image.IsParent(context.TODO()) if err != nil { return call.ReplyErrorOccurred(err.Error()) } @@ -503,7 +504,7 @@ func (i *LibpodAPI) DeleteUnusedImages(call iopodman.VarlinkCall) error { return call.ReplyErrorOccurred(err.Error()) } if len(containers) == 0 { - if err := img.Remove(false); err != nil { + if err := img.Remove(context.TODO(), false); err != nil { return call.ReplyErrorOccurred(err.Error()) } deletedImages = append(deletedImages, img.ID()) @@ -739,7 +740,7 @@ func (i *LibpodAPI) ContainerRunlabel(call iopodman.VarlinkCall, input iopodman. // ImagesPrune .... func (i *LibpodAPI) ImagesPrune(call iopodman.VarlinkCall, all bool) error { - prunedImages, err := i.Runtime.ImageRuntime().PruneImages(all) + prunedImages, err := i.Runtime.ImageRuntime().PruneImages(context.TODO(), all) if err != nil { return call.ReplyErrorOccurred(err.Error()) } |