vbatts/registry
1
0
Fork 0
Code Issues Pull requests Releases Wiki Activity

Refactor Blob Service API

Browse source 
This PR refactors the blob service API to be oriented around blob descriptors.
Identified by digests, blobs become an abstract entity that can be read and
written using a descriptor as a handle. This allows blobs to take many forms,
such as a ReadSeekCloser or a simple byte buffer, allowing blob oriented
operations to better integrate with blob agnostic APIs (such as the `io`
package). The error definitions are now better organized to reflect conditions
that can only be seen when interacting with the blob API.

The main benefit of this is to separate the much smaller metadata from large
file storage. Many benefits also follow from this. Reading and writing has
been separated into discrete services. Backend implementation is also
simplified, by reducing the amount of metadata that needs to be picked up to
simply serve a read. This also improves cacheability.

"Opening" a blob simply consists of an access check (Stat) and a path
calculation. Caching is greatly simplified and we've made the mapping of
provisional to canonical hashes a first-class concept. BlobDescriptorService
and BlobProvider can be combined in different ways to achieve varying effects.

Recommend Review Approach
-------------------------

This is a very large patch. While apologies are in order, we are getting a
considerable amount of refactoring. Most changes follow from the changes to
the root package (distribution), so start there. From there, the main changes
are in storage. Looking at (*repository).Blobs will help to understand the how
the linkedBlobStore is wired. One can explore the internals within and also
branch out into understanding the changes to the caching layer. Following the
descriptions below will also help to guide you.

To reduce the chances for regressions, it was critical that major changes to
unit tests were avoided. Where possible, they are left untouched and where
not, the spirit is hopefully captured. Pay particular attention to where
behavior may have changed.

Storage
-------

The primary changes to the `storage` package, other than the interface
updates, were to merge the layerstore and blobstore. Blob access is now
layered even further. The first layer, blobStore, exposes a global
`BlobStatter` and `BlobProvider`. Operations here provide a fast path for most
read operations that don't take access control into account. The
`linkedBlobStore` layers on top of the `blobStore`, providing repository-
scoped blob link management in the backend. The `linkedBlobStore` implements
the full `BlobStore` suite, providing access-controlled, repository-local blob
writers. The abstraction between the two is slightly broken in that
`linkedBlobStore` is the only channel under which one can write into the global
blob store. The `linkedBlobStore` also provides flexibility in that it can act
over different link sets depending on configuration. This allows us to use the
same code for signature links, manifest links and blob links.  Eventually, we
will fully consolidate this storage.

The improved cache flow comes from the `linkedBlobStatter` component
of `linkedBlobStore`. Using a `cachedBlobStatter`, these combine together to
provide a simple cache hierarchy that should streamline access checks on read
and write operations, or at least provide a single path to optimize. The
metrics have been changed in a slightly incompatible way since the former
operations, Fetch and Exists, are no longer relevant.

The fileWriter and fileReader have been slightly modified to support the rest
of the changes. The most interesting is the removal of the `Stat` call from
`newFileReader`. This was the source of unnecessary round trips that were only
present to look up the size of the resulting reader. Now, one must simply pass
in the size, requiring the caller to decide whether or not the `Stat` call is
appropriate. In several cases, it turned out the caller already had the size
already. The `WriterAt` implementation has been removed from `fileWriter`,
since it is no longer required for `BlobWriter`, reducing the number of paths
which writes may take.

Cache
-----

Unfortunately, the `cache` package required a near full rewrite. It was pretty
mechanical in that the cache is oriented around the `BlobDescriptorService`
slightly modified to include the ability to set the values for individual
digests. While the implementation is oriented towards caching, it can act as a
primary store. Provisions are in place to have repository local metadata, in
addition to global metadata. Fallback is implemented as a part of the storage
package to maintain this flexibility.

One unfortunate side-effect is that caching is now repository-scoped, rather
than global. This should have little effect on performance but may increase
memory usage.

Handlers
--------

The `handlers` package has been updated to leverage the new API. For the most
part, the changes are superficial or mechanical based on the API changes. This
did expose a bug in the handling of provisional vs canonical digests that was
fixed in the unit tests.

Configuration
-------------

One user-facing change has been made to the configuration and is updated in
the associated documentation. The `layerinfo` cache parameter has been
deprecated by the `blobdescriptor` cache parameter. Both are equivalent and
configuration files should be backward compatible.

Notifications
-------------

Changes the `notification` package are simply to support the interface
changes.

Context
-------

A small change has been made to the tracing log-level. Traces have been moved
from "info" to "debug" level to reduce output when not needed.

Signed-off-by: Stephen J Day <stephen.day@docker.com>
This commit is contained in:
Stephen J Day 2015-05-12 00:10:29 -07:00
parent dc348d720b
commit 08401cfdd6
44 changed files with 2426 additions and 2270 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

21
docs/handlers/api_test.go
View file

@ -93,8 +93,8 @@ func TestURLPrefix(t *testing.T) {
}
// TestLayerAPI conducts a full test of the of the layer api.
func TestLayerAPI(t *testing.T) {
// TestBlobAPI conducts a full test of the of the blob api.
func TestBlobAPI(t *testing.T) {
// TODO(stevvooe): This test code is complete junk but it should cover the
// complete flow. This must be broken down and checked against the
// specification *before* we submit the final to docker core.
@ -213,6 +213,13 @@ func TestLayerAPI(t *testing.T) {
// Now, push just a chunk
layerFile.Seek(0, 0)
canonicalDigester := digest.NewCanonicalDigester()
if _, err := io.Copy(canonicalDigester, layerFile); err != nil {
t.Fatalf("error copying to digest: %v", err)
}
canonicalDigest := canonicalDigester.Digest()
layerFile.Seek(0, 0)
uploadURLBase, uploadUUID = startPushLayer(t, env.builder, imageName)
uploadURLBase, dgst := pushChunk(t, env.builder, imageName, uploadURLBase, layerFile, layerLength)
finishUpload(t, env.builder, imageName, uploadURLBase, dgst)
@ -226,7 +233,7 @@ func TestLayerAPI(t *testing.T) {
checkResponse(t, "checking head on existing layer", resp, http.StatusOK)
checkHeaders(t, resp, http.Header{
"Content-Length": []string{fmt.Sprint(layerLength)},
"Docker-Content-Digest": []string{layerDigest.String()},
"Docker-Content-Digest": []string{canonicalDigest.String()},
})
// ----------------
@ -239,7 +246,7 @@ func TestLayerAPI(t *testing.T) {
checkResponse(t, "fetching layer", resp, http.StatusOK)
checkHeaders(t, resp, http.Header{
"Content-Length": []string{fmt.Sprint(layerLength)},
"Docker-Content-Digest": []string{layerDigest.String()},
"Docker-Content-Digest": []string{canonicalDigest.String()},
})
// Verify the body
@ -272,9 +279,9 @@ func TestLayerAPI(t *testing.T) {
checkResponse(t, "fetching layer", resp, http.StatusOK)
checkHeaders(t, resp, http.Header{
"Content-Length": []string{fmt.Sprint(layerLength)},
"Docker-Content-Digest": []string{layerDigest.String()},
"ETag": []string{layerDigest.String()},
"Cache-Control": []string{"max-age=86400"},
"Docker-Content-Digest": []string{canonicalDigest.String()},
"ETag": []string{canonicalDigest.String()},
"Cache-Control": []string{"max-age=31536000"},
})
// Matching etag, gives 304

24
docs/handlers/app.go
View file

@ -67,9 +67,9 @@ func NewApp(ctx context.Context, configuration configuration.Configuration) *App
})
app.register(v2.RouteNameManifest, imageManifestDispatcher)
app.register(v2.RouteNameTags, tagsDispatcher)
app.register(v2.RouteNameBlob, layerDispatcher)
app.register(v2.RouteNameBlobUpload, layerUploadDispatcher)
app.register(v2.RouteNameBlobUploadChunk, layerUploadDispatcher)
app.register(v2.RouteNameBlob, blobDispatcher)
app.register(v2.RouteNameBlobUpload, blobUploadDispatcher)
app.register(v2.RouteNameBlobUploadChunk, blobUploadDispatcher)
var err error
app.driver, err = factory.Create(configuration.Storage.Type(), configuration.Storage.Parameters())
@ -103,18 +103,24 @@ func NewApp(ctx context.Context, configuration configuration.Configuration) *App
// configure storage caches
if cc, ok := configuration.Storage["cache"]; ok {
switch cc["layerinfo"] {
v, ok := cc["blobdescriptor"]
if !ok {
// Backwards compatible: "layerinfo" == "blobdescriptor"
v = cc["layerinfo"]
}
switch v {
case "redis":
if app.redis == nil {
panic("redis configuration required to use for layerinfo cache")
}
app.registry = storage.NewRegistryWithDriver(app, app.driver, cache.NewRedisLayerInfoCache(app.redis))
ctxu.GetLogger(app).Infof("using redis layerinfo cache")
app.registry = storage.NewRegistryWithDriver(app, app.driver, cache.NewRedisBlobDescriptorCacheProvider(app.redis))
ctxu.GetLogger(app).Infof("using redis blob descriptor cache")
case "inmemory":
app.registry = storage.NewRegistryWithDriver(app, app.driver, cache.NewInMemoryLayerInfoCache())
ctxu.GetLogger(app).Infof("using inmemory layerinfo cache")
app.registry = storage.NewRegistryWithDriver(app, app.driver, cache.NewInMemoryBlobDescriptorCacheProvider())
ctxu.GetLogger(app).Infof("using inmemory blob descriptor cache")
default:
if cc["layerinfo"] != "" {
if v != "" {
ctxu.GetLogger(app).Warnf("unkown cache type %q, caching disabled", configuration.Storage["cache"])
}
}

2
docs/handlers/app_test.go
View file

@ -30,7 +30,7 @@ func TestAppDispatcher(t *testing.T) {
Context: ctx,
router: v2.Router(),
driver: driver,
registry: storage.NewRegistryWithDriver(ctx, driver, cache.NewInMemoryLayerInfoCache()),
registry: storage.NewRegistryWithDriver(ctx, driver, cache.NewInMemoryBlobDescriptorCacheProvider()),
}
server := httptest.NewServer(app)
router := v2.Router()

69
docs/handlers/blob.go Normal file
View file

@ -0,0 +1,69 @@
package handlers
import (
"net/http"
"github.com/docker/distribution"
"github.com/docker/distribution/context"
"github.com/docker/distribution/digest"
"github.com/docker/distribution/registry/api/v2"
"github.com/gorilla/handlers"
)
// blobDispatcher uses the request context to build a blobHandler.
func blobDispatcher(ctx *Context, r *http.Request) http.Handler {
dgst, err := getDigest(ctx)
if err != nil {
if err == errDigestNotAvailable {
return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
w.WriteHeader(http.StatusNotFound)
ctx.Errors.Push(v2.ErrorCodeDigestInvalid, err)
})
}
return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
ctx.Errors.Push(v2.ErrorCodeDigestInvalid, err)
})
}
blobHandler := &blobHandler{
Context: ctx,
Digest: dgst,
}
return handlers.MethodHandler{
"GET": http.HandlerFunc(blobHandler.GetBlob),
"HEAD": http.HandlerFunc(blobHandler.GetBlob),
}
}
// blobHandler serves http blob requests.
type blobHandler struct {
*Context
Digest digest.Digest
}
// GetBlob fetches the binary data from backend storage returns it in the
// response.
func (bh *blobHandler) GetBlob(w http.ResponseWriter, r *http.Request) {
context.GetLogger(bh).Debug("GetBlob")
blobs := bh.Repository.Blobs(bh)
desc, err := blobs.Stat(bh, bh.Digest)
if err != nil {
if err == distribution.ErrBlobUnknown {
w.WriteHeader(http.StatusNotFound)
bh.Errors.Push(v2.ErrorCodeBlobUnknown, bh.Digest)
} else {
bh.Errors.Push(v2.ErrorCodeUnknown, err)
}
return
}
if err := blobs.ServeBlob(bh, w, r, desc.Digest); err != nil {
context.GetLogger(bh).Debugf("unexpected error getting blob HTTP handler: %v", err)
bh.Errors.Push(v2.ErrorCodeUnknown, err)
return
}
}

355
docs/handlers/blobupload.go Normal file
View file

@ -0,0 +1,355 @@
package handlers
import (
"fmt"
"io"
"net/http"
"net/url"
"os"
"github.com/docker/distribution"
ctxu "github.com/docker/distribution/context"
"github.com/docker/distribution/digest"
"github.com/docker/distribution/registry/api/v2"
"github.com/gorilla/handlers"
)
// blobUploadDispatcher constructs and returns the blob upload handler for the
// given request context.
func blobUploadDispatcher(ctx *Context, r *http.Request) http.Handler {
buh := &blobUploadHandler{
Context: ctx,
UUID: getUploadUUID(ctx),
}
handler := http.Handler(handlers.MethodHandler{
"POST": http.HandlerFunc(buh.StartBlobUpload),
"GET": http.HandlerFunc(buh.GetUploadStatus),
"HEAD": http.HandlerFunc(buh.GetUploadStatus),
"PATCH": http.HandlerFunc(buh.PatchBlobData),
"PUT": http.HandlerFunc(buh.PutBlobUploadComplete),
"DELETE": http.HandlerFunc(buh.CancelBlobUpload),
})
if buh.UUID != "" {
state, err := hmacKey(ctx.Config.HTTP.Secret).unpackUploadState(r.FormValue("_state"))
if err != nil {
return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
ctxu.GetLogger(ctx).Infof("error resolving upload: %v", err)
w.WriteHeader(http.StatusBadRequest)
buh.Errors.Push(v2.ErrorCodeBlobUploadInvalid, err)
})
}
buh.State = state
if state.Name != ctx.Repository.Name() {
return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
ctxu.GetLogger(ctx).Infof("mismatched repository name in upload state: %q != %q", state.Name, buh.Repository.Name())
w.WriteHeader(http.StatusBadRequest)
buh.Errors.Push(v2.ErrorCodeBlobUploadInvalid, err)
})
}
if state.UUID != buh.UUID {
return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
ctxu.GetLogger(ctx).Infof("mismatched uuid in upload state: %q != %q", state.UUID, buh.UUID)
w.WriteHeader(http.StatusBadRequest)
buh.Errors.Push(v2.ErrorCodeBlobUploadInvalid, err)
})
}
blobs := ctx.Repository.Blobs(buh)
upload, err := blobs.Resume(buh, buh.UUID)
if err != nil {
ctxu.GetLogger(ctx).Errorf("error resolving upload: %v", err)
if err == distribution.ErrBlobUploadUnknown {
return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
w.WriteHeader(http.StatusNotFound)
buh.Errors.Push(v2.ErrorCodeBlobUploadUnknown, err)
})
}
return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
w.WriteHeader(http.StatusInternalServerError)
buh.Errors.Push(v2.ErrorCodeUnknown, err)
})
}
buh.Upload = upload
if state.Offset > 0 {
// Seek the blob upload to the correct spot if it's non-zero.
// These error conditions should be rare and demonstrate really
// problems. We basically cancel the upload and tell the client to
// start over.
if nn, err := upload.Seek(buh.State.Offset, os.SEEK_SET); err != nil {
defer upload.Close()
ctxu.GetLogger(ctx).Infof("error seeking blob upload: %v", err)
return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
w.WriteHeader(http.StatusBadRequest)
buh.Errors.Push(v2.ErrorCodeBlobUploadInvalid, err)
upload.Cancel(buh)
})
} else if nn != buh.State.Offset {
defer upload.Close()
ctxu.GetLogger(ctx).Infof("seek to wrong offest: %d != %d", nn, buh.State.Offset)
return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
w.WriteHeader(http.StatusBadRequest)
buh.Errors.Push(v2.ErrorCodeBlobUploadInvalid, err)
upload.Cancel(buh)
})
}
}
handler = closeResources(handler, buh.Upload)
}
return handler
}
// blobUploadHandler handles the http blob upload process.
type blobUploadHandler struct {
*Context
// UUID identifies the upload instance for the current request. Using UUID
// to key blob writers since this implementation uses UUIDs.
UUID string
Upload distribution.BlobWriter
State blobUploadState
}
// StartBlobUpload begins the blob upload process and allocates a server-side
// blob writer session.
func (buh *blobUploadHandler) StartBlobUpload(w http.ResponseWriter, r *http.Request) {
blobs := buh.Repository.Blobs(buh)
upload, err := blobs.Create(buh)
if err != nil {
w.WriteHeader(http.StatusInternalServerError) // Error conditions here?
buh.Errors.Push(v2.ErrorCodeUnknown, err)
return
}
buh.Upload = upload
defer buh.Upload.Close()
if err := buh.blobUploadResponse(w, r, true); err != nil {
w.WriteHeader(http.StatusInternalServerError) // Error conditions here?
buh.Errors.Push(v2.ErrorCodeUnknown, err)
return
}
w.Header().Set("Docker-Upload-UUID", buh.Upload.ID())
w.WriteHeader(http.StatusAccepted)
}
// GetUploadStatus returns the status of a given upload, identified by id.
func (buh *blobUploadHandler) GetUploadStatus(w http.ResponseWriter, r *http.Request) {
if buh.Upload == nil {
w.WriteHeader(http.StatusNotFound)
buh.Errors.Push(v2.ErrorCodeBlobUploadUnknown)
return
}
// TODO(dmcgowan): Set last argument to false in blobUploadResponse when
// resumable upload is supported. This will enable returning a non-zero
// range for clients to begin uploading at an offset.
if err := buh.blobUploadResponse(w, r, true); err != nil {
w.WriteHeader(http.StatusInternalServerError) // Error conditions here?
buh.Errors.Push(v2.ErrorCodeUnknown, err)
return
}
w.Header().Set("Docker-Upload-UUID", buh.UUID)
w.WriteHeader(http.StatusNoContent)
}
// PatchBlobData writes data to an upload.
func (buh *blobUploadHandler) PatchBlobData(w http.ResponseWriter, r *http.Request) {
if buh.Upload == nil {
w.WriteHeader(http.StatusNotFound)
buh.Errors.Push(v2.ErrorCodeBlobUploadUnknown)
return
}
ct := r.Header.Get("Content-Type")
if ct != "" && ct != "application/octet-stream" {
w.WriteHeader(http.StatusBadRequest)
// TODO(dmcgowan): encode error
return
}
// TODO(dmcgowan): support Content-Range header to seek and write range
// Copy the data
if _, err := io.Copy(buh.Upload, r.Body); err != nil {
ctxu.GetLogger(buh).Errorf("unknown error copying into upload: %v", err)
w.WriteHeader(http.StatusInternalServerError)
buh.Errors.Push(v2.ErrorCodeUnknown, err)
return
}
if err := buh.blobUploadResponse(w, r, false); err != nil {
w.WriteHeader(http.StatusInternalServerError) // Error conditions here?
buh.Errors.Push(v2.ErrorCodeUnknown, err)
return
}
w.WriteHeader(http.StatusAccepted)
}
// PutBlobUploadComplete takes the final request of a blob upload. The
// request may include all the blob data or no blob data. Any data
// provided is received and verified. If successful, the blob is linked
// into the blob store and 201 Created is returned with the canonical
// url of the blob.
func (buh *blobUploadHandler) PutBlobUploadComplete(w http.ResponseWriter, r *http.Request) {
if buh.Upload == nil {
w.WriteHeader(http.StatusNotFound)
buh.Errors.Push(v2.ErrorCodeBlobUploadUnknown)
return
}
dgstStr := r.FormValue("digest") // TODO(stevvooe): Support multiple digest parameters!
if dgstStr == "" {
// no digest? return error, but allow retry.
w.WriteHeader(http.StatusBadRequest)
buh.Errors.Push(v2.ErrorCodeDigestInvalid, "digest missing")
return
}
dgst, err := digest.ParseDigest(dgstStr)
if err != nil {
// no digest? return error, but allow retry.
w.WriteHeader(http.StatusNotFound)
buh.Errors.Push(v2.ErrorCodeDigestInvalid, "digest parsing failed")
return
}
// Read in the data, if any.
if _, err := io.Copy(buh.Upload, r.Body); err != nil {
ctxu.GetLogger(buh).Errorf("unknown error copying into upload: %v", err)
w.WriteHeader(http.StatusInternalServerError)
buh.Errors.Push(v2.ErrorCodeUnknown, err)
return
}
desc, err := buh.Upload.Commit(buh, distribution.Descriptor{
Digest: dgst,
// TODO(stevvooe): This isn't wildly important yet, but we should
// really set the length and mediatype. For now, we can let the
// backend take care of this.
})
if err != nil {
switch err := err.(type) {
case distribution.ErrBlobInvalidDigest:
w.WriteHeader(http.StatusBadRequest)
buh.Errors.Push(v2.ErrorCodeDigestInvalid, err)
default:
switch err {
case distribution.ErrBlobInvalidLength, distribution.ErrBlobDigestUnsupported:
w.WriteHeader(http.StatusBadRequest)
buh.Errors.Push(v2.ErrorCodeBlobUploadInvalid, err)
default:
ctxu.GetLogger(buh).Errorf("unknown error completing upload: %#v", err)
w.WriteHeader(http.StatusInternalServerError)
buh.Errors.Push(v2.ErrorCodeUnknown, err)
}
}
// Clean up the backend blob data if there was an error.
if err := buh.Upload.Cancel(buh); err != nil {
// If the cleanup fails, all we can do is observe and report.
ctxu.GetLogger(buh).Errorf("error canceling upload after error: %v", err)
}
return
}
// Build our canonical blob url
blobURL, err := buh.urlBuilder.BuildBlobURL(buh.Repository.Name(), desc.Digest)
if err != nil {
buh.Errors.Push(v2.ErrorCodeUnknown, err)
w.WriteHeader(http.StatusInternalServerError)
return
}
w.Header().Set("Location", blobURL)
w.Header().Set("Content-Length", "0")
w.Header().Set("Docker-Content-Digest", desc.Digest.String())
w.WriteHeader(http.StatusCreated)
}
// CancelBlobUpload cancels an in-progress upload of a blob.
func (buh *blobUploadHandler) CancelBlobUpload(w http.ResponseWriter, r *http.Request) {
if buh.Upload == nil {
w.WriteHeader(http.StatusNotFound)
buh.Errors.Push(v2.ErrorCodeBlobUploadUnknown)
return
}
w.Header().Set("Docker-Upload-UUID", buh.UUID)
if err := buh.Upload.Cancel(buh); err != nil {
ctxu.GetLogger(buh).Errorf("error encountered canceling upload: %v", err)
w.WriteHeader(http.StatusInternalServerError)
buh.Errors.PushErr(err)
}
w.WriteHeader(http.StatusNoContent)
}
// blobUploadResponse provides a standard request for uploading blobs and
// chunk responses. This sets the correct headers but the response status is
// left to the caller. The fresh argument is used to ensure that new blob
// uploads always start at a 0 offset. This allows disabling resumable push by
// always returning a 0 offset on check status.
func (buh *blobUploadHandler) blobUploadResponse(w http.ResponseWriter, r *http.Request, fresh bool) error {
var offset int64
if !fresh {
var err error
offset, err = buh.Upload.Seek(0, os.SEEK_CUR)
if err != nil {
ctxu.GetLogger(buh).Errorf("unable get current offset of blob upload: %v", err)
return err
}
}
// TODO(stevvooe): Need a better way to manage the upload state automatically.
buh.State.Name = buh.Repository.Name()
buh.State.UUID = buh.Upload.ID()
buh.State.Offset = offset
buh.State.StartedAt = buh.Upload.StartedAt()
token, err := hmacKey(buh.Config.HTTP.Secret).packUploadState(buh.State)
if err != nil {
ctxu.GetLogger(buh).Infof("error building upload state token: %s", err)
return err
}
uploadURL, err := buh.urlBuilder.BuildBlobUploadChunkURL(
buh.Repository.Name(), buh.Upload.ID(),
url.Values{
"_state": []string{token},
})
if err != nil {
ctxu.GetLogger(buh).Infof("error building upload url: %s", err)
return err
}
endRange := offset
if endRange > 0 {
endRange = endRange - 1
}
w.Header().Set("Docker-Upload-UUID", buh.UUID)
w.Header().Set("Location", uploadURL)
w.Header().Set("Content-Length", "0")
w.Header().Set("Range", fmt.Sprintf("0-%d", endRange))
return nil
}

14
docs/handlers/hmac.go
View file

@ -9,9 +9,9 @@ import (
"time"
)
// layerUploadState captures the state serializable state of the layer upload.
type layerUploadState struct {
// name is the primary repository under which the layer will be linked.
// blobUploadState captures the state serializable state of the blob upload.
type blobUploadState struct {
// name is the primary repository under which the blob will be linked.
Name string
// UUID identifies the upload.
@ -26,10 +26,10 @@ type layerUploadState struct {
type hmacKey string
// unpackUploadState unpacks and validates the layer upload state from the
// unpackUploadState unpacks and validates the blob upload state from the
// token, using the hmacKey secret.
func (secret hmacKey) unpackUploadState(token string) (layerUploadState, error) {
var state layerUploadState
func (secret hmacKey) unpackUploadState(token string) (blobUploadState, error) {
var state blobUploadState
tokenBytes, err := base64.URLEncoding.DecodeString(token)
if err != nil {
@ -59,7 +59,7 @@ func (secret hmacKey) unpackUploadState(token string) (layerUploadState, error)
// packUploadState packs the upload state signed with and hmac digest using
// the hmacKey secret, encoding to url safe base64. The resulting token can be
// used to share data with minimized risk of external tampering.
func (secret hmacKey) packUploadState(lus layerUploadState) (string, error) {
func (secret hmacKey) packUploadState(lus blobUploadState) (string, error) {
mac := hmac.New(sha256.New, []byte(secret))
p, err := json.Marshal(lus)
if err != nil {

12
docs/handlers/hmac_test.go
View file

@ -2,7 +2,7 @@ package handlers
import "testing"
var layerUploadStates = []layerUploadState{
var blobUploadStates = []blobUploadState{
{
Name: "hello",
UUID: "abcd-1234-qwer-0987",
@ -45,7 +45,7 @@ var secrets = []string{
func TestLayerUploadTokens(t *testing.T) {
secret := hmacKey("supersecret")
for _, testcase := range layerUploadStates {
for _, testcase := range blobUploadStates {
token, err := secret.packUploadState(testcase)
if err != nil {
t.Fatal(err)
@ -56,7 +56,7 @@ func TestLayerUploadTokens(t *testing.T) {
t.Fatal(err)
}
assertLayerUploadStateEquals(t, testcase, lus)
assertBlobUploadStateEquals(t, testcase, lus)
}
}
@ -68,7 +68,7 @@ func TestHMACValidation(t *testing.T) {
secret2 := hmacKey(secret)
badSecret := hmacKey("DifferentSecret")
for _, testcase := range layerUploadStates {
for _, testcase := range blobUploadStates {
token, err := secret1.packUploadState(testcase)
if err != nil {
t.Fatal(err)
@ -79,7 +79,7 @@ func TestHMACValidation(t *testing.T) {
t.Fatal(err)
}
assertLayerUploadStateEquals(t, testcase, lus)
assertBlobUploadStateEquals(t, testcase, lus)
_, err = badSecret.unpackUploadState(token)
if err == nil {
@ -104,7 +104,7 @@ func TestHMACValidation(t *testing.T) {
}
}
func assertLayerUploadStateEquals(t *testing.T, expected layerUploadState, received layerUploadState) {
func assertBlobUploadStateEquals(t *testing.T, expected blobUploadState, received blobUploadState) {
if expected.Name != received.Name {
t.Fatalf("Expected Name=%q, Received Name=%q", expected.Name, received.Name)
}

6
docs/handlers/images.go
View file

@ -136,14 +136,12 @@ func (imh *imageManifestHandler) PutImageManifest(w http.ResponseWriter, r *http
case distribution.ErrManifestVerification:
for _, verificationError := range err {
switch verificationError := verificationError.(type) {
case distribution.ErrUnknownLayer:
imh.Errors.Push(v2.ErrorCodeBlobUnknown, verificationError.FSLayer)
case distribution.ErrManifestBlobUnknown:
imh.Errors.Push(v2.ErrorCodeBlobUnknown, verificationError.Digest)
case distribution.ErrManifestUnverified:
imh.Errors.Push(v2.ErrorCodeManifestUnverified)
default:
if verificationError == digest.ErrDigestInvalidFormat {
// TODO(stevvooe): We need to really need to move all
// errors to types. Its much more straightforward.
imh.Errors.Push(v2.ErrorCodeDigestInvalid)
} else {
imh.Errors.PushErr(verificationError)

74
docs/handlers/layer.go
View file

@ -1,74 +0,0 @@
package handlers
import (
"net/http"
"github.com/docker/distribution"
"github.com/docker/distribution/context"
"github.com/docker/distribution/digest"
"github.com/docker/distribution/registry/api/v2"
"github.com/gorilla/handlers"
)
// layerDispatcher uses the request context to build a layerHandler.
func layerDispatcher(ctx *Context, r *http.Request) http.Handler {
dgst, err := getDigest(ctx)
if err != nil {
if err == errDigestNotAvailable {
return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
w.WriteHeader(http.StatusNotFound)
ctx.Errors.Push(v2.ErrorCodeDigestInvalid, err)
})
}
return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
ctx.Errors.Push(v2.ErrorCodeDigestInvalid, err)
})
}
layerHandler := &layerHandler{
Context: ctx,
Digest: dgst,
}
return handlers.MethodHandler{
"GET": http.HandlerFunc(layerHandler.GetLayer),
"HEAD": http.HandlerFunc(layerHandler.GetLayer),
}
}
// layerHandler serves http layer requests.
type layerHandler struct {
*Context
Digest digest.Digest
}
// GetLayer fetches the binary data from backend storage returns it in the
// response.
func (lh *layerHandler) GetLayer(w http.ResponseWriter, r *http.Request) {
context.GetLogger(lh).Debug("GetImageLayer")
layers := lh.Repository.Layers()
layer, err := layers.Fetch(lh.Digest)
if err != nil {
switch err := err.(type) {
case distribution.ErrUnknownLayer:
w.WriteHeader(http.StatusNotFound)
lh.Errors.Push(v2.ErrorCodeBlobUnknown, err.FSLayer)
default:
lh.Errors.Push(v2.ErrorCodeUnknown, err)
}
return
}
handler, err := layer.Handler(r)
if err != nil {
context.GetLogger(lh).Debugf("unexpected error getting layer HTTP handler: %s", err)
lh.Errors.Push(v2.ErrorCodeUnknown, err)
return
}
handler.ServeHTTP(w, r)
}

344
docs/handlers/layerupload.go
View file

@ -1,344 +0,0 @@
package handlers
import (
"fmt"
"io"
"net/http"
"net/url"
"os"
"github.com/docker/distribution"
ctxu "github.com/docker/distribution/context"
"github.com/docker/distribution/digest"
"github.com/docker/distribution/registry/api/v2"
"github.com/gorilla/handlers"
)
// layerUploadDispatcher constructs and returns the layer upload handler for
// the given request context.
func layerUploadDispatcher(ctx *Context, r *http.Request) http.Handler {
luh := &layerUploadHandler{
Context: ctx,
UUID: getUploadUUID(ctx),
}
handler := http.Handler(handlers.MethodHandler{
"POST": http.HandlerFunc(luh.StartLayerUpload),
"GET": http.HandlerFunc(luh.GetUploadStatus),
"HEAD": http.HandlerFunc(luh.GetUploadStatus),
"PATCH": http.HandlerFunc(luh.PatchLayerData),
"PUT": http.HandlerFunc(luh.PutLayerUploadComplete),
"DELETE": http.HandlerFunc(luh.CancelLayerUpload),
})
if luh.UUID != "" {
state, err := hmacKey(ctx.Config.HTTP.Secret).unpackUploadState(r.FormValue("_state"))
if err != nil {
return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
ctxu.GetLogger(ctx).Infof("error resolving upload: %v", err)
w.WriteHeader(http.StatusBadRequest)
luh.Errors.Push(v2.ErrorCodeBlobUploadInvalid, err)
})
}
luh.State = state
if state.Name != ctx.Repository.Name() {
return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
ctxu.GetLogger(ctx).Infof("mismatched repository name in upload state: %q != %q", state.Name, luh.Repository.Name())
w.WriteHeader(http.StatusBadRequest)
luh.Errors.Push(v2.ErrorCodeBlobUploadInvalid, err)
})
}
if state.UUID != luh.UUID {
return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
ctxu.GetLogger(ctx).Infof("mismatched uuid in upload state: %q != %q", state.UUID, luh.UUID)
w.WriteHeader(http.StatusBadRequest)
luh.Errors.Push(v2.ErrorCodeBlobUploadInvalid, err)
})
}
layers := ctx.Repository.Layers()
upload, err := layers.Resume(luh.UUID)
if err != nil {
ctxu.GetLogger(ctx).Errorf("error resolving upload: %v", err)
if err == distribution.ErrLayerUploadUnknown {
return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
w.WriteHeader(http.StatusNotFound)
luh.Errors.Push(v2.ErrorCodeBlobUploadUnknown, err)
})
}
return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
w.WriteHeader(http.StatusInternalServerError)
luh.Errors.Push(v2.ErrorCodeUnknown, err)
})
}
luh.Upload = upload
if state.Offset > 0 {
// Seek the layer upload to the correct spot if it's non-zero.
// These error conditions should be rare and demonstrate really
// problems. We basically cancel the upload and tell the client to
// start over.
if nn, err := upload.Seek(luh.State.Offset, os.SEEK_SET); err != nil {
defer upload.Close()
ctxu.GetLogger(ctx).Infof("error seeking layer upload: %v", err)
return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
w.WriteHeader(http.StatusBadRequest)
luh.Errors.Push(v2.ErrorCodeBlobUploadInvalid, err)
upload.Cancel()
})
} else if nn != luh.State.Offset {
defer upload.Close()
ctxu.GetLogger(ctx).Infof("seek to wrong offest: %d != %d", nn, luh.State.Offset)
return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
w.WriteHeader(http.StatusBadRequest)
luh.Errors.Push(v2.ErrorCodeBlobUploadInvalid, err)
upload.Cancel()
})
}
}
handler = closeResources(handler, luh.Upload)
}
return handler
}
// layerUploadHandler handles the http layer upload process.
type layerUploadHandler struct {
*Context
// UUID identifies the upload instance for the current request.
UUID string
Upload distribution.LayerUpload
State layerUploadState
}
// StartLayerUpload begins the layer upload process and allocates a server-
// side upload session.
func (luh *layerUploadHandler) StartLayerUpload(w http.ResponseWriter, r *http.Request) {
layers := luh.Repository.Layers()
upload, err := layers.Upload()
if err != nil {
w.WriteHeader(http.StatusInternalServerError) // Error conditions here?
luh.Errors.Push(v2.ErrorCodeUnknown, err)
return
}
luh.Upload = upload
defer luh.Upload.Close()
if err := luh.layerUploadResponse(w, r, true); err != nil {
w.WriteHeader(http.StatusInternalServerError) // Error conditions here?
luh.Errors.Push(v2.ErrorCodeUnknown, err)
return
}
w.Header().Set("Docker-Upload-UUID", luh.Upload.UUID())
w.WriteHeader(http.StatusAccepted)
}
// GetUploadStatus returns the status of a given upload, identified by uuid.
func (luh *layerUploadHandler) GetUploadStatus(w http.ResponseWriter, r *http.Request) {
if luh.Upload == nil {
w.WriteHeader(http.StatusNotFound)
luh.Errors.Push(v2.ErrorCodeBlobUploadUnknown)
return
}
// TODO(dmcgowan): Set last argument to false in layerUploadResponse when
// resumable upload is supported. This will enable returning a non-zero
// range for clients to begin uploading at an offset.
if err := luh.layerUploadResponse(w, r, true); err != nil {
w.WriteHeader(http.StatusInternalServerError) // Error conditions here?
luh.Errors.Push(v2.ErrorCodeUnknown, err)
return
}
w.Header().Set("Docker-Upload-UUID", luh.UUID)
w.WriteHeader(http.StatusNoContent)
}
// PatchLayerData writes data to an upload.
func (luh *layerUploadHandler) PatchLayerData(w http.ResponseWriter, r *http.Request) {
if luh.Upload == nil {
w.WriteHeader(http.StatusNotFound)
luh.Errors.Push(v2.ErrorCodeBlobUploadUnknown)
return
}
ct := r.Header.Get("Content-Type")
if ct != "" && ct != "application/octet-stream" {
w.WriteHeader(http.StatusBadRequest)
// TODO(dmcgowan): encode error
return
}
// TODO(dmcgowan): support Content-Range header to seek and write range
// Copy the data
if _, err := io.Copy(luh.Upload, r.Body); err != nil {
ctxu.GetLogger(luh).Errorf("unknown error copying into upload: %v", err)
w.WriteHeader(http.StatusInternalServerError)
luh.Errors.Push(v2.ErrorCodeUnknown, err)
return
}
if err := luh.layerUploadResponse(w, r, false); err != nil {
w.WriteHeader(http.StatusInternalServerError) // Error conditions here?
luh.Errors.Push(v2.ErrorCodeUnknown, err)
return
}
w.WriteHeader(http.StatusAccepted)
}
// PutLayerUploadComplete takes the final request of a layer upload. The
// request may include all the layer data or no layer data. Any data
// provided is received and verified. If successful, the layer is linked
// into the blob store and 201 Created is returned with the canonical
// url of the layer.
func (luh *layerUploadHandler) PutLayerUploadComplete(w http.ResponseWriter, r *http.Request) {
if luh.Upload == nil {
w.WriteHeader(http.StatusNotFound)
luh.Errors.Push(v2.ErrorCodeBlobUploadUnknown)
return
}
dgstStr := r.FormValue("digest") // TODO(stevvooe): Support multiple digest parameters!
if dgstStr == "" {
// no digest? return error, but allow retry.
w.WriteHeader(http.StatusBadRequest)
luh.Errors.Push(v2.ErrorCodeDigestInvalid, "digest missing")
return
}
dgst, err := digest.ParseDigest(dgstStr)
if err != nil {
// no digest? return error, but allow retry.
w.WriteHeader(http.StatusNotFound)
luh.Errors.Push(v2.ErrorCodeDigestInvalid, "digest parsing failed")
return
}
// TODO(stevvooe): Consider checking the error on this copy.
// Theoretically, problems should be detected during verification but we
// may miss a root cause.
// Read in the data, if any.
if _, err := io.Copy(luh.Upload, r.Body); err != nil {
ctxu.GetLogger(luh).Errorf("unknown error copying into upload: %v", err)
w.WriteHeader(http.StatusInternalServerError)
luh.Errors.Push(v2.ErrorCodeUnknown, err)
return
}
layer, err := luh.Upload.Finish(dgst)
if err != nil {
switch err := err.(type) {
case distribution.ErrLayerInvalidDigest:
w.WriteHeader(http.StatusBadRequest)
luh.Errors.Push(v2.ErrorCodeDigestInvalid, err)
default:
ctxu.GetLogger(luh).Errorf("unknown error completing upload: %#v", err)
w.WriteHeader(http.StatusInternalServerError)
luh.Errors.Push(v2.ErrorCodeUnknown, err)
}
// Clean up the backend layer data if there was an error.
if err := luh.Upload.Cancel(); err != nil {
// If the cleanup fails, all we can do is observe and report.
ctxu.GetLogger(luh).Errorf("error canceling upload after error: %v", err)
}
return
}
// Build our canonical layer url
layerURL, err := luh.urlBuilder.BuildBlobURL(luh.Repository.Name(), layer.Digest())
if err != nil {
luh.Errors.Push(v2.ErrorCodeUnknown, err)
w.WriteHeader(http.StatusInternalServerError)
return
}
w.Header().Set("Location", layerURL)
w.Header().Set("Content-Length", "0")
w.Header().Set("Docker-Content-Digest", layer.Digest().String())
w.WriteHeader(http.StatusCreated)
}
// CancelLayerUpload cancels an in-progress upload of a layer.
func (luh *layerUploadHandler) CancelLayerUpload(w http.ResponseWriter, r *http.Request) {
if luh.Upload == nil {
w.WriteHeader(http.StatusNotFound)
luh.Errors.Push(v2.ErrorCodeBlobUploadUnknown)
return
}
w.Header().Set("Docker-Upload-UUID", luh.UUID)
if err := luh.Upload.Cancel(); err != nil {
ctxu.GetLogger(luh).Errorf("error encountered canceling upload: %v", err)
w.WriteHeader(http.StatusInternalServerError)
luh.Errors.PushErr(err)
}
w.WriteHeader(http.StatusNoContent)
}
// layerUploadResponse provides a standard request for uploading layers and
// chunk responses. This sets the correct headers but the response status is
// left to the caller. The fresh argument is used to ensure that new layer
// uploads always start at a 0 offset. This allows disabling resumable push
// by always returning a 0 offset on check status.
func (luh *layerUploadHandler) layerUploadResponse(w http.ResponseWriter, r *http.Request, fresh bool) error {
var offset int64
if !fresh {
var err error
offset, err = luh.Upload.Seek(0, os.SEEK_CUR)
if err != nil {
ctxu.GetLogger(luh).Errorf("unable get current offset of layer upload: %v", err)
return err
}
}
// TODO(stevvooe): Need a better way to manage the upload state automatically.
luh.State.Name = luh.Repository.Name()
luh.State.UUID = luh.Upload.UUID()
luh.State.Offset = offset
luh.State.StartedAt = luh.Upload.StartedAt()
token, err := hmacKey(luh.Config.HTTP.Secret).packUploadState(luh.State)
if err != nil {
ctxu.GetLogger(luh).Infof("error building upload state token: %s", err)
return err
}
uploadURL, err := luh.urlBuilder.BuildBlobUploadChunkURL(
luh.Repository.Name(), luh.Upload.UUID(),
url.Values{
"_state": []string{token},
})
if err != nil {
ctxu.GetLogger(luh).Infof("error building upload url: %s", err)
return err
}
endRange := offset
if endRange > 0 {
endRange = endRange - 1
}
w.Header().Set("Docker-Upload-UUID", luh.UUID)
w.Header().Set("Location", uploadURL)
w.Header().Set("Content-Length", "0")
w.Header().Set("Range", fmt.Sprintf("0-%d", endRange))
return nil
}

225
docs/storage/layer_test.go → docs/storage/blob_test.go
View file

@ -13,14 +13,13 @@ import (
"github.com/docker/distribution/context"
"github.com/docker/distribution/digest"
"github.com/docker/distribution/registry/storage/cache"
storagedriver "github.com/docker/distribution/registry/storage/driver"
"github.com/docker/distribution/registry/storage/driver/inmemory"
"github.com/docker/distribution/testutil"
)
// TestSimpleLayerUpload covers the layer upload process, exercising common
// TestSimpleBlobUpload covers the blob upload process, exercising common
// error paths that might be seen during an upload.
func TestSimpleLayerUpload(t *testing.T) {
func TestSimpleBlobUpload(t *testing.T) {
randomDataReader, tarSumStr, err := testutil.CreateRandomTarFile()
if err != nil {
@ -36,35 +35,35 @@ func TestSimpleLayerUpload(t *testing.T) {
ctx := context.Background()
imageName := "foo/bar"
driver := inmemory.New()
registry := NewRegistryWithDriver(ctx, driver, cache.NewInMemoryLayerInfoCache())
registry := NewRegistryWithDriver(ctx, driver, cache.NewInMemoryBlobDescriptorCacheProvider())
repository, err := registry.Repository(ctx, imageName)
if err != nil {
t.Fatalf("unexpected error getting repo: %v", err)
}
ls := repository.Layers()
bs := repository.Blobs(ctx)
h := sha256.New()
rd := io.TeeReader(randomDataReader, h)
layerUpload, err := ls.Upload()
blobUpload, err := bs.Create(ctx)
if err != nil {
t.Fatalf("unexpected error starting layer upload: %s", err)
}
// Cancel the upload then restart it
if err := layerUpload.Cancel(); err != nil {
if err := blobUpload.Cancel(ctx); err != nil {
t.Fatalf("unexpected error during upload cancellation: %v", err)
}
// Do a resume, get unknown upload
layerUpload, err = ls.Resume(layerUpload.UUID())
if err != distribution.ErrLayerUploadUnknown {
blobUpload, err = bs.Resume(ctx, blobUpload.ID())
if err != distribution.ErrBlobUploadUnknown {
t.Fatalf("unexpected error resuming upload, should be unkown: %v", err)
}
// Restart!
layerUpload, err = ls.Upload()
blobUpload, err = bs.Create(ctx)
if err != nil {
t.Fatalf("unexpected error starting layer upload: %s", err)
}
@ -75,7 +74,7 @@ func TestSimpleLayerUpload(t *testing.T) {
t.Fatalf("error getting seeker size of random data: %v", err)
}
nn, err := io.Copy(layerUpload, rd)
nn, err := io.Copy(blobUpload, rd)
if err != nil {
t.Fatalf("unexpected error uploading layer data: %v", err)
}
@ -84,46 +83,51 @@ func TestSimpleLayerUpload(t *testing.T) {
t.Fatalf("layer data write incomplete")
}
offset, err := layerUpload.Seek(0, os.SEEK_CUR)
offset, err := blobUpload.Seek(0, os.SEEK_CUR)
if err != nil {
t.Fatalf("unexpected error seeking layer upload: %v", err)
}
if offset != nn {
t.Fatalf("layerUpload not updated with correct offset: %v != %v", offset, nn)
t.Fatalf("blobUpload not updated with correct offset: %v != %v", offset, nn)
}
layerUpload.Close()
blobUpload.Close()
// Do a resume, for good fun
layerUpload, err = ls.Resume(layerUpload.UUID())
blobUpload, err = bs.Resume(ctx, blobUpload.ID())
if err != nil {
t.Fatalf("unexpected error resuming upload: %v", err)
}
sha256Digest := digest.NewDigest("sha256", h)
layer, err := layerUpload.Finish(dgst)
desc, err := blobUpload.Commit(ctx, distribution.Descriptor{Digest: dgst})
if err != nil {
t.Fatalf("unexpected error finishing layer upload: %v", err)
}
// After finishing an upload, it should no longer exist.
if _, err := ls.Resume(layerUpload.UUID()); err != distribution.ErrLayerUploadUnknown {
if _, err := bs.Resume(ctx, blobUpload.ID()); err != distribution.ErrBlobUploadUnknown {
t.Fatalf("expected layer upload to be unknown, got %v", err)
}
// Test for existence.
exists, err := ls.Exists(layer.Digest())
statDesc, err := bs.Stat(ctx, desc.Digest)
if err != nil {
t.Fatalf("unexpected error checking for existence: %v", err)
t.Fatalf("unexpected error checking for existence: %v, %#v", err, bs)
}
if !exists {
t.Fatalf("layer should now exist")
if statDesc != desc {
t.Fatalf("descriptors not equal: %v != %v", statDesc, desc)
}
rc, err := bs.Open(ctx, desc.Digest)
if err != nil {
t.Fatalf("unexpected error opening blob for read: %v", err)
}
defer rc.Close()
h.Reset()
nn, err = io.Copy(h, layer)
nn, err = io.Copy(h, rc)
if err != nil {
t.Fatalf("error reading layer: %v", err)
}
@ -137,21 +141,21 @@ func TestSimpleLayerUpload(t *testing.T) {
}
}
// TestSimpleLayerRead just creates a simple layer file and ensures that basic
// TestSimpleBlobRead just creates a simple blob file and ensures that basic
// open, read, seek, read works. More specific edge cases should be covered in
// other tests.
func TestSimpleLayerRead(t *testing.T) {
func TestSimpleBlobRead(t *testing.T) {
ctx := context.Background()
imageName := "foo/bar"
driver := inmemory.New()
registry := NewRegistryWithDriver(ctx, driver, cache.NewInMemoryLayerInfoCache())
registry := NewRegistryWithDriver(ctx, driver, cache.NewInMemoryBlobDescriptorCacheProvider())
repository, err := registry.Repository(ctx, imageName)
if err != nil {
t.Fatalf("unexpected error getting repo: %v", err)
}
ls := repository.Layers()
bs := repository.Blobs(ctx)
randomLayerReader, tarSumStr, err := testutil.CreateRandomTarFile()
randomLayerReader, tarSumStr, err := testutil.CreateRandomTarFile() // TODO(stevvooe): Consider using just a random string.
if err != nil {
t.Fatalf("error creating random data: %v", err)
}
@ -159,31 +163,14 @@ func TestSimpleLayerRead(t *testing.T) {
dgst := digest.Digest(tarSumStr)
// Test for existence.
exists, err := ls.Exists(dgst)
if err != nil {
t.Fatalf("unexpected error checking for existence: %v", err)
desc, err := bs.Stat(ctx, dgst)
if err != distribution.ErrBlobUnknown {
t.Fatalf("expected not found error when testing for existence: %v", err)
}
if exists {
t.Fatalf("layer should not exist")
}
// Try to get the layer and make sure we get a not found error
layer, err := ls.Fetch(dgst)
if err == nil {
t.Fatalf("error expected fetching unknown layer")
}
switch err.(type) {
case distribution.ErrUnknownLayer:
err = nil
default:
t.Fatalf("unexpected error fetching non-existent layer: %v", err)
}
randomLayerDigest, err := writeTestLayer(driver, defaultPathMapper, imageName, dgst, randomLayerReader)
if err != nil {
t.Fatalf("unexpected error writing test layer: %v", err)
rc, err := bs.Open(ctx, dgst)
if err != distribution.ErrBlobUnknown {
t.Fatalf("expected not found error when opening non-existent blob: %v", err)
}
randomLayerSize, err := seekerSize(randomLayerReader)
@ -191,45 +178,57 @@ func TestSimpleLayerRead(t *testing.T) {
t.Fatalf("error getting seeker size for random layer: %v", err)
}
layer, err = ls.Fetch(dgst)
descBefore := distribution.Descriptor{Digest: dgst, MediaType: "application/octet-stream", Length: randomLayerSize}
t.Logf("desc: %v", descBefore)
desc, err = addBlob(ctx, bs, descBefore, randomLayerReader)
if err != nil {
t.Fatal(err)
t.Fatalf("error adding blob to blobservice: %v", err)
}
defer layer.Close()
if desc.Length != randomLayerSize {
t.Fatalf("committed blob has incorrect length: %v != %v", desc.Length, randomLayerSize)
}
rc, err = bs.Open(ctx, desc.Digest) // note that we are opening with original digest.
if err != nil {
t.Fatalf("error opening blob with %v: %v", dgst, err)
}
defer rc.Close()
// Now check the sha digest and ensure its the same
h := sha256.New()
nn, err := io.Copy(h, layer)
if err != nil && err != io.EOF {
nn, err := io.Copy(h, rc)
if err != nil {
t.Fatalf("unexpected error copying to hash: %v", err)
}
if nn != randomLayerSize {
t.Fatalf("stored incorrect number of bytes in layer: %d != %d", nn, randomLayerSize)
t.Fatalf("stored incorrect number of bytes in blob: %d != %d", nn, randomLayerSize)
}
sha256Digest := digest.NewDigest("sha256", h)
if sha256Digest != randomLayerDigest {
t.Fatalf("fetched digest does not match: %q != %q", sha256Digest, randomLayerDigest)
if sha256Digest != desc.Digest {
t.Fatalf("fetched digest does not match: %q != %q", sha256Digest, desc.Digest)
}
// Now seek back the layer, read the whole thing and check against randomLayerData
offset, err := layer.Seek(0, os.SEEK_SET)
// Now seek back the blob, read the whole thing and check against randomLayerData
offset, err := rc.Seek(0, os.SEEK_SET)
if err != nil {
t.Fatalf("error seeking layer: %v", err)
t.Fatalf("error seeking blob: %v", err)
}
if offset != 0 {
t.Fatalf("seek failed: expected 0 offset, got %d", offset)
}
p, err := ioutil.ReadAll(layer)
p, err := ioutil.ReadAll(rc)
if err != nil {
t.Fatalf("error reading all of layer: %v", err)
t.Fatalf("error reading all of blob: %v", err)
}
if len(p) != int(randomLayerSize) {
t.Fatalf("layer data read has different length: %v != %v", len(p), randomLayerSize)
t.Fatalf("blob data read has different length: %v != %v", len(p), randomLayerSize)
}
// Reset the randomLayerReader and read back the buffer
@ -253,19 +252,26 @@ func TestLayerUploadZeroLength(t *testing.T) {
ctx := context.Background()
imageName := "foo/bar"
driver := inmemory.New()
registry := NewRegistryWithDriver(ctx, driver, cache.NewInMemoryLayerInfoCache())
registry := NewRegistryWithDriver(ctx, driver, cache.NewInMemoryBlobDescriptorCacheProvider())
repository, err := registry.Repository(ctx, imageName)
if err != nil {
t.Fatalf("unexpected error getting repo: %v", err)
}
ls := repository.Layers()
bs := repository.Blobs(ctx)
upload, err := ls.Upload()
wr, err := bs.Create(ctx)
if err != nil {
t.Fatalf("unexpected error starting upload: %v", err)
}
io.Copy(upload, bytes.NewReader([]byte{}))
nn, err := io.Copy(wr, bytes.NewReader([]byte{}))
if err != nil {
t.Fatalf("error copying into blob writer: %v", err)
}
if nn != 0 {
t.Fatalf("unexpected number of bytes copied: %v > 0", nn)
}
dgst, err := digest.FromReader(bytes.NewReader([]byte{}))
if err != nil {
@ -277,37 +283,16 @@ func TestLayerUploadZeroLength(t *testing.T) {
t.Fatalf("digest not as expected: %v != %v", dgst, digest.DigestTarSumV1EmptyTar)
}
layer, err := upload.Finish(dgst)
desc, err := wr.Commit(ctx, distribution.Descriptor{Digest: dgst})
if err != nil {
t.Fatalf("unexpected error finishing upload: %v", err)
t.Fatalf("unexpected error committing write: %v", err)
}
if layer.Digest() != dgst {
t.Fatalf("unexpected digest: %v != %v", layer.Digest(), dgst)
if desc.Digest != dgst {
t.Fatalf("unexpected digest: %v != %v", desc.Digest, dgst)
}
}
// writeRandomLayer creates a random layer under name and tarSum using driver
// and pathMapper. An io.ReadSeeker with the data is returned, along with the
// sha256 hex digest.
func writeRandomLayer(driver storagedriver.StorageDriver, pathMapper *pathMapper, name string) (rs io.ReadSeeker, tarSum digest.Digest, sha256digest digest.Digest, err error) {
reader, tarSumStr, err := testutil.CreateRandomTarFile()
if err != nil {
return nil, "", "", err
}
tarSum = digest.Digest(tarSumStr)
// Now, actually create the layer.
randomLayerDigest, err := writeTestLayer(driver, pathMapper, name, tarSum, ioutil.NopCloser(reader))
if _, err := reader.Seek(0, os.SEEK_SET); err != nil {
return nil, "", "", err
}
return reader, tarSum, randomLayerDigest, err
}
// seekerSize seeks to the end of seeker, checks the size and returns it to
// the original state, returning the size. The state of the seeker should be
// treated as unknown if an error is returned.
@ -334,46 +319,20 @@ func seekerSize(seeker io.ReadSeeker) (int64, error) {
return end, nil
}
// createTestLayer creates a simple test layer in the provided driver under
// tarsum dgst, returning the sha256 digest location. This is implemented
// piecemeal and should probably be replaced by the uploader when it's ready.
func writeTestLayer(driver storagedriver.StorageDriver, pathMapper *pathMapper, name string, dgst digest.Digest, content io.Reader) (digest.Digest, error) {
h := sha256.New()
rd := io.TeeReader(content, h)
p, err := ioutil.ReadAll(rd)
// addBlob simply consumes the reader and inserts into the blob service,
// returning a descriptor on success.
func addBlob(ctx context.Context, bs distribution.BlobIngester, desc distribution.Descriptor, rd io.Reader) (distribution.Descriptor, error) {
wr, err := bs.Create(ctx)
if err != nil {
return "", nil
return distribution.Descriptor{}, err
}
defer wr.Cancel(ctx)
if nn, err := io.Copy(wr, rd); err != nil {
return distribution.Descriptor{}, err
} else if nn != desc.Length {
return distribution.Descriptor{}, fmt.Errorf("incorrect number of bytes copied: %v != %v", nn, desc.Length)
}
blobDigestSHA := digest.NewDigest("sha256", h)
blobPath, err := pathMapper.path(blobDataPathSpec{
digest: dgst,
})
ctx := context.Background()
if err := driver.PutContent(ctx, blobPath, p); err != nil {
return "", err
}
if err != nil {
return "", err
}
layerLinkPath, err := pathMapper.path(layerLinkPathSpec{
name: name,
digest: dgst,
})
if err != nil {
return "", err
}
if err := driver.PutContent(ctx, layerLinkPath, []byte(dgst)); err != nil {
return "", nil
}
return blobDigestSHA, err
return wr.Commit(ctx, desc)
}

72
docs/storage/blobserver.go Normal file
View file

@ -0,0 +1,72 @@
package storage
import (
"fmt"
"net/http"
"time"
"github.com/docker/distribution"
"github.com/docker/distribution/context"
"github.com/docker/distribution/digest"
"github.com/docker/distribution/registry/storage/driver"
)
// TODO(stevvooe): This should configurable in the future.
const blobCacheControlMaxAge = 365 * 24 * time.Hour
// blobServer simply serves blobs from a driver instance using a path function
// to identify paths and a descriptor service to fill in metadata.
type blobServer struct {
driver driver.StorageDriver
statter distribution.BlobStatter
pathFn func(dgst digest.Digest) (string, error)
}
func (bs *blobServer) ServeBlob(ctx context.Context, w http.ResponseWriter, r *http.Request, dgst digest.Digest) error {
desc, err := bs.statter.Stat(ctx, dgst)
if err != nil {
return err
}
path, err := bs.pathFn(desc.Digest)
if err != nil {
return err
}
redirectURL, err := bs.driver.URLFor(ctx, path, map[string]interface{}{"method": r.Method})
switch err {
case nil:
// Redirect to storage URL.
http.Redirect(w, r, redirectURL, http.StatusTemporaryRedirect)
case driver.ErrUnsupportedMethod:
// Fallback to serving the content directly.
br, err := newFileReader(ctx, bs.driver, path, desc.Length)
if err != nil {
return err
}
defer br.Close()
w.Header().Set("ETag", desc.Digest.String()) // If-None-Match handled by ServeContent
w.Header().Set("Cache-Control", fmt.Sprintf("max-age=%.f", blobCacheControlMaxAge.Seconds()))
if w.Header().Get("Docker-Content-Digest") == "" {
w.Header().Set("Docker-Content-Digest", desc.Digest.String())
}
if w.Header().Get("Content-Type") == "" {
// Set the content type if not already set.
w.Header().Set("Content-Type", desc.MediaType)
}
if w.Header().Get("Content-Length") == "" {
// Set the content length if not already set.
w.Header().Set("Content-Length", fmt.Sprint(desc.Length))
}
http.ServeContent(w, r, desc.Digest.String(), time.Time{}, br)
}
// Some unexpected error.
return err
}

221
docs/storage/blobstore.go
View file

@ -1,133 +1,94 @@
package storage
import (
"fmt"
"github.com/docker/distribution"
"github.com/docker/distribution/context"
"github.com/docker/distribution/digest"
storagedriver "github.com/docker/distribution/registry/storage/driver"
"github.com/docker/distribution/registry/storage/driver"
)
// TODO(stevvooe): Currently, the blobStore implementation used by the
// manifest store. The layer store should be refactored to better leverage the
// blobStore, reducing duplicated code.
// blobStore implements a generalized blob store over a driver, supporting the
// read side and link management. This object is intentionally a leaky
// abstraction, providing utility methods that support creating and traversing
// backend links.
// blobStore implements a the read side of the blob store interface over a
// driver without enforcing per-repository membership. This object is
// intentionally a leaky abstraction, providing utility methods that support
// creating and traversing backend links.
type blobStore struct {
driver storagedriver.StorageDriver
driver driver.StorageDriver
pm *pathMapper
ctx context.Context
statter distribution.BlobStatter
}
// exists reports whether or not the path exists. If the driver returns error
// other than storagedriver.PathNotFound, an error may be returned.
func (bs *blobStore) exists(dgst digest.Digest) (bool, error) {
path, err := bs.path(dgst)
var _ distribution.BlobProvider = &blobStore{}
if err != nil {
return false, err
}
ok, err := exists(bs.ctx, bs.driver, path)
if err != nil {
return false, err
}
return ok, nil
}
// get retrieves the blob by digest, returning it a byte slice. This should
// only be used for small objects.
func (bs *blobStore) get(dgst digest.Digest) ([]byte, error) {
// Get implements the BlobReadService.Get call.
func (bs *blobStore) Get(ctx context.Context, dgst digest.Digest) ([]byte, error) {
bp, err := bs.path(dgst)
if err != nil {
return nil, err
}
return bs.driver.GetContent(bs.ctx, bp)
p, err := bs.driver.GetContent(ctx, bp)
if err != nil {
switch err.(type) {
case driver.PathNotFoundError:
return nil, distribution.ErrBlobUnknown
}
// link links the path to the provided digest by writing the digest into the
// target file.
func (bs *blobStore) link(path string, dgst digest.Digest) error {
if exists, err := bs.exists(dgst); err != nil {
return err
} else if !exists {
return fmt.Errorf("cannot link non-existent blob")
return nil, err
}
// The contents of the "link" file are the exact string contents of the
// digest, which is specified in that package.
return bs.driver.PutContent(bs.ctx, path, []byte(dgst))
return p, err
}
// linked reads the link at path and returns the content.
func (bs *blobStore) linked(path string) ([]byte, error) {
linked, err := bs.readlink(path)
func (bs *blobStore) Open(ctx context.Context, dgst digest.Digest) (distribution.ReadSeekCloser, error) {
desc, err := bs.statter.Stat(ctx, dgst)
if err != nil {
return nil, err
}
return bs.get(linked)
}
// readlink returns the linked digest at path.
func (bs *blobStore) readlink(path string) (digest.Digest, error) {
content, err := bs.driver.GetContent(bs.ctx, path)
path, err := bs.path(desc.Digest)
if err != nil {
return "", err
return nil, err
}
linked, err := digest.ParseDigest(string(content))
if err != nil {
return "", err
return newFileReader(ctx, bs.driver, path, desc.Length)
}
if exists, err := bs.exists(linked); err != nil {
return "", err
} else if !exists {
return "", fmt.Errorf("link %q invalid: blob %s does not exist", path, linked)
}
return linked, nil
}
// resolve reads the digest link at path and returns the blob store link.
func (bs *blobStore) resolve(path string) (string, error) {
dgst, err := bs.readlink(path)
if err != nil {
return "", err
}
return bs.path(dgst)
}
// put stores the content p in the blob store, calculating the digest. If the
// Put stores the content p in the blob store, calculating the digest. If the
// content is already present, only the digest will be returned. This should
// only be used for small objects, such as manifests.
func (bs *blobStore) put(p []byte) (digest.Digest, error) {
// only be used for small objects, such as manifests. This implemented as a convenience for other Put implementations
func (bs *blobStore) Put(ctx context.Context, mediaType string, p []byte) (distribution.Descriptor, error) {
dgst, err := digest.FromBytes(p)
if err != nil {
context.GetLogger(bs.ctx).Errorf("error digesting content: %v, %s", err, string(p))
return "", err
context.GetLogger(ctx).Errorf("blobStore: error digesting content: %v, %s", err, string(p))
return distribution.Descriptor{}, err
}
desc, err := bs.statter.Stat(ctx, dgst)
if err == nil {
// content already present
return desc, nil
} else if err != distribution.ErrBlobUnknown {
context.GetLogger(ctx).Errorf("blobStore: error stating content (%v): %#v", dgst, err)
// real error, return it
return distribution.Descriptor{}, err
}
bp, err := bs.path(dgst)
if err != nil {
return "", err
return distribution.Descriptor{}, err
}
// If the content already exists, just return the digest.
if exists, err := bs.exists(dgst); err != nil {
return "", err
} else if exists {
return dgst, nil
}
// TODO(stevvooe): Write out mediatype here, as well.
return dgst, bs.driver.PutContent(bs.ctx, bp, p)
return distribution.Descriptor{
Length: int64(len(p)),
// NOTE(stevvooe): The central blob store firewalls media types from
// other users. The caller should look this up and override the value
// for the specific repository.
MediaType: "application/octet-stream",
Digest: dgst,
}, bs.driver.PutContent(ctx, bp, p)
}
// path returns the canonical path for the blob identified by digest. The blob
@ -144,16 +105,86 @@ func (bs *blobStore) path(dgst digest.Digest) (string, error) {
return bp, nil
}
// exists provides a utility method to test whether or not a path exists
func exists(ctx context.Context, driver storagedriver.StorageDriver, path string) (bool, error) {
if _, err := driver.Stat(ctx, path); err != nil {
// link links the path to the provided digest by writing the digest into the
// target file. Caller must ensure that the blob actually exists.
func (bs *blobStore) link(ctx context.Context, path string, dgst digest.Digest) error {
// The contents of the "link" file are the exact string contents of the
// digest, which is specified in that package.
return bs.driver.PutContent(ctx, path, []byte(dgst))
}
// readlink returns the linked digest at path.
func (bs *blobStore) readlink(ctx context.Context, path string) (digest.Digest, error) {
content, err := bs.driver.GetContent(ctx, path)
if err != nil {
return "", err
}
linked, err := digest.ParseDigest(string(content))
if err != nil {
return "", err
}
return linked, nil
}
// resolve reads the digest link at path and returns the blob store path.
func (bs *blobStore) resolve(ctx context.Context, path string) (string, error) {
dgst, err := bs.readlink(ctx, path)
if err != nil {
return "", err
}
return bs.path(dgst)
}
type blobStatter struct {
driver driver.StorageDriver
pm *pathMapper
}
var _ distribution.BlobStatter = &blobStatter{}
// Stat implements BlobStatter.Stat by returning the descriptor for the blob
// in the main blob store. If this method returns successfully, there is
// strong guarantee that the blob exists and is available.
func (bs *blobStatter) Stat(ctx context.Context, dgst digest.Digest) (distribution.Descriptor, error) {
path, err := bs.pm.path(blobDataPathSpec{
digest: dgst,
})
if err != nil {
return distribution.Descriptor{}, err
}
fi, err := bs.driver.Stat(ctx, path)
if err != nil {
switch err := err.(type) {
case storagedriver.PathNotFoundError:
return false, nil
case driver.PathNotFoundError:
return distribution.Descriptor{}, distribution.ErrBlobUnknown
default:
return false, err
return distribution.Descriptor{}, err
}
}
return true, nil
if fi.IsDir() {
// NOTE(stevvooe): This represents a corruption situation. Somehow, we
// calculated a blob path and then detected a directory. We log the
// error and then error on the side of not knowing about the blob.
context.GetLogger(ctx).Warnf("blob path should not be a directory: %q", path)
return distribution.Descriptor{}, distribution.ErrBlobUnknown
}
// TODO(stevvooe): Add method to resolve the mediatype. We can store and
// cache a "global" media type for the blob, even if a specific repo has a
// mediatype that overrides the main one.
return distribution.Descriptor{
Length: fi.Size(),
// NOTE(stevvooe): The central blob store firewalls media types from
// other users. The caller should look this up and override the value
// for the specific repository.
MediaType: "application/octet-stream",
Digest: dgst,
}, nil
}

469
docs/storage/blobwriter.go Normal file
View file

@ -0,0 +1,469 @@
package storage
import (
"fmt"
"io"
"os"
"path"
"strconv"
"time"
"github.com/Sirupsen/logrus"
"github.com/docker/distribution"
"github.com/docker/distribution/context"
"github.com/docker/distribution/digest"
storagedriver "github.com/docker/distribution/registry/storage/driver"
)
// layerWriter is used to control the various aspects of resumable
// layer upload. It implements the LayerUpload interface.
type blobWriter struct {
blobStore *linkedBlobStore
id string
startedAt time.Time
resumableDigester digest.ResumableDigester
// implementes io.WriteSeeker, io.ReaderFrom and io.Closer to satisfy
// LayerUpload Interface
bufferedFileWriter
}
var _ distribution.BlobWriter = &blobWriter{}
// ID returns the identifier for this upload.
func (bw *blobWriter) ID() string {
return bw.id
}
func (bw *blobWriter) StartedAt() time.Time {
return bw.startedAt
}
// Commit marks the upload as completed, returning a valid descriptor. The
// final size and digest are checked against the first descriptor provided.
func (bw *blobWriter) Commit(ctx context.Context, desc distribution.Descriptor) (distribution.Descriptor, error) {
context.GetLogger(ctx).Debug("(*blobWriter).Commit")
if err := bw.bufferedFileWriter.Close(); err != nil {
return distribution.Descriptor{}, err
}
canonical, err := bw.validateBlob(ctx, desc)
if err != nil {
return distribution.Descriptor{}, err
}
if err := bw.moveBlob(ctx, canonical); err != nil {
return distribution.Descriptor{}, err
}
if err := bw.blobStore.linkBlob(ctx, canonical, desc.Digest); err != nil {
return distribution.Descriptor{}, err
}
if err := bw.removeResources(ctx); err != nil {
return distribution.Descriptor{}, err
}
return canonical, nil
}
// Rollback the blob upload process, releasing any resources associated with
// the writer and canceling the operation.
func (bw *blobWriter) Cancel(ctx context.Context) error {
context.GetLogger(ctx).Debug("(*blobWriter).Rollback")
if err := bw.removeResources(ctx); err != nil {
return err
}
bw.Close()
return nil
}
func (bw *blobWriter) Write(p []byte) (int, error) {
if bw.resumableDigester == nil {
return bw.bufferedFileWriter.Write(p)
}
// Ensure that the current write offset matches how many bytes have been
// written to the digester. If not, we need to update the digest state to
// match the current write position.
if err := bw.resumeHashAt(bw.blobStore.ctx, bw.offset); err != nil {
return 0, err
}
return io.MultiWriter(&bw.bufferedFileWriter, bw.resumableDigester).Write(p)
}
func (bw *blobWriter) ReadFrom(r io.Reader) (n int64, err error) {
if bw.resumableDigester == nil {
return bw.bufferedFileWriter.ReadFrom(r)
}
// Ensure that the current write offset matches how many bytes have been
// written to the digester. If not, we need to update the digest state to
// match the current write position.
if err := bw.resumeHashAt(bw.blobStore.ctx, bw.offset); err != nil {
return 0, err
}
return bw.bufferedFileWriter.ReadFrom(io.TeeReader(r, bw.resumableDigester))
}
func (bw *blobWriter) Close() error {
if bw.err != nil {
return bw.err
}
if bw.resumableDigester != nil {
if err := bw.storeHashState(bw.blobStore.ctx); err != nil {
return err
}
}
return bw.bufferedFileWriter.Close()
}
// validateBlob checks the data against the digest, returning an error if it
// does not match. The canonical descriptor is returned.
func (bw *blobWriter) validateBlob(ctx context.Context, desc distribution.Descriptor) (distribution.Descriptor, error) {
var (
verified, fullHash bool
canonical digest.Digest
)
if desc.Digest == "" {
// if no descriptors are provided, we have nothing to validate
// against. We don't really want to support this for the registry.
return distribution.Descriptor{}, distribution.ErrBlobInvalidDigest{
Reason: fmt.Errorf("cannot validate against empty digest"),
}
}
// Stat the on disk file
if fi, err := bw.bufferedFileWriter.driver.Stat(ctx, bw.path); err != nil {
switch err := err.(type) {
case storagedriver.PathNotFoundError:
// NOTE(stevvooe): We really don't care if the file is
// not actually present for the reader. We now assume
// that the desc length is zero.
desc.Length = 0
default:
// Any other error we want propagated up the stack.
return distribution.Descriptor{}, err
}
} else {
if fi.IsDir() {
return distribution.Descriptor{}, fmt.Errorf("unexpected directory at upload location %q", bw.path)
}
bw.size = fi.Size()
}
if desc.Length > 0 {
if desc.Length != bw.size {
return distribution.Descriptor{}, distribution.ErrBlobInvalidLength
}
} else {
// if provided 0 or negative length, we can assume caller doesn't know or
// care about length.
desc.Length = bw.size
}
if bw.resumableDigester != nil {
// Restore the hasher state to the end of the upload.
if err := bw.resumeHashAt(ctx, bw.size); err != nil {
return distribution.Descriptor{}, err
}
canonical = bw.resumableDigester.Digest()
if canonical.Algorithm() == desc.Digest.Algorithm() {
// Common case: client and server prefer the same canonical digest
// algorithm - currently SHA256.
verified = desc.Digest == canonical
} else {
// The client wants to use a different digest algorithm. They'll just
// have to be patient and wait for us to download and re-hash the
// uploaded content using that digest algorithm.
fullHash = true
}
} else {
// Not using resumable digests, so we need to hash the entire layer.
fullHash = true
}
if fullHash {
digester := digest.NewCanonicalDigester()
digestVerifier, err := digest.NewDigestVerifier(desc.Digest)
if err != nil {
return distribution.Descriptor{}, err
}
// Read the file from the backend driver and validate it.
fr, err := newFileReader(ctx, bw.bufferedFileWriter.driver, bw.path, desc.Length)
if err != nil {
return distribution.Descriptor{}, err
}
tr := io.TeeReader(fr, digester)
if _, err := io.Copy(digestVerifier, tr); err != nil {
return distribution.Descriptor{}, err
}
canonical = digester.Digest()
verified = digestVerifier.Verified()
}
if !verified {
context.GetLoggerWithFields(ctx,
map[string]interface{}{
"canonical": canonical,
"provided": desc.Digest,
}, "canonical", "provided").
Errorf("canonical digest does match provided digest")
return distribution.Descriptor{}, distribution.ErrBlobInvalidDigest{
Digest: desc.Digest,
Reason: fmt.Errorf("content does not match digest"),
}
}
// update desc with canonical hash
desc.Digest = canonical
if desc.MediaType == "" {
desc.MediaType = "application/octet-stream"
}
return desc, nil
}
// moveBlob moves the data into its final, hash-qualified destination,
// identified by dgst. The layer should be validated before commencing the
// move.
func (bw *blobWriter) moveBlob(ctx context.Context, desc distribution.Descriptor) error {
blobPath, err := bw.blobStore.pm.path(blobDataPathSpec{
digest: desc.Digest,
})
if err != nil {
return err
}
// Check for existence
if _, err := bw.blobStore.driver.Stat(ctx, blobPath); err != nil {
switch err := err.(type) {
case storagedriver.PathNotFoundError:
break // ensure that it doesn't exist.
default:
return err
}
} else {
// If the path exists, we can assume that the content has already
// been uploaded, since the blob storage is content-addressable.
// While it may be corrupted, detection of such corruption belongs
// elsewhere.
return nil
}
// If no data was received, we may not actually have a file on disk. Check
// the size here and write a zero-length file to blobPath if this is the
// case. For the most part, this should only ever happen with zero-length
// tars.
if _, err := bw.blobStore.driver.Stat(ctx, bw.path); err != nil {
switch err := err.(type) {
case storagedriver.PathNotFoundError:
// HACK(stevvooe): This is slightly dangerous: if we verify above,
// get a hash, then the underlying file is deleted, we risk moving
// a zero-length blob into a nonzero-length blob location. To
// prevent this horrid thing, we employ the hack of only allowing
// to this happen for the zero tarsum.
if desc.Digest == digest.DigestSha256EmptyTar {
return bw.blobStore.driver.PutContent(ctx, blobPath, []byte{})
}
// We let this fail during the move below.
logrus.
WithField("upload.id", bw.ID()).
WithField("digest", desc.Digest).Warnf("attempted to move zero-length content with non-zero digest")
default:
return err // unrelated error
}
}
// TODO(stevvooe): We should also write the mediatype when executing this move.
return bw.blobStore.driver.Move(ctx, bw.path, blobPath)
}
type hashStateEntry struct {
offset int64
path string
}
// getStoredHashStates returns a slice of hashStateEntries for this upload.
func (bw *blobWriter) getStoredHashStates(ctx context.Context) ([]hashStateEntry, error) {
uploadHashStatePathPrefix, err := bw.blobStore.pm.path(uploadHashStatePathSpec{
name: bw.blobStore.repository.Name(),
id: bw.id,
alg: bw.resumableDigester.Digest().Algorithm(),
list: true,
})
if err != nil {
return nil, err
}
paths, err := bw.blobStore.driver.List(ctx, uploadHashStatePathPrefix)
if err != nil {
if _, ok := err.(storagedriver.PathNotFoundError); !ok {
return nil, err
}
// Treat PathNotFoundError as no entries.
paths = nil
}
hashStateEntries := make([]hashStateEntry, 0, len(paths))
for _, p := range paths {
pathSuffix := path.Base(p)
// The suffix should be the offset.
offset, err := strconv.ParseInt(pathSuffix, 0, 64)
if err != nil {
logrus.Errorf("unable to parse offset from upload state path %q: %s", p, err)
}
hashStateEntries = append(hashStateEntries, hashStateEntry{offset: offset, path: p})
}
return hashStateEntries, nil
}
// resumeHashAt attempts to restore the state of the internal hash function
// by loading the most recent saved hash state less than or equal to the given
// offset. Any unhashed bytes remaining less than the given offset are hashed
// from the content uploaded so far.
func (bw *blobWriter) resumeHashAt(ctx context.Context, offset int64) error {
if offset < 0 {
return fmt.Errorf("cannot resume hash at negative offset: %d", offset)
}
if offset == int64(bw.resumableDigester.Len()) {
// State of digester is already at the requested offset.
return nil
}
// List hash states from storage backend.
var hashStateMatch hashStateEntry
hashStates, err := bw.getStoredHashStates(ctx)
if err != nil {
return fmt.Errorf("unable to get stored hash states with offset %d: %s", offset, err)
}
// Find the highest stored hashState with offset less than or equal to
// the requested offset.
for _, hashState := range hashStates {
if hashState.offset == offset {
hashStateMatch = hashState
break // Found an exact offset match.
} else if hashState.offset < offset && hashState.offset > hashStateMatch.offset {
// This offset is closer to the requested offset.
hashStateMatch = hashState
} else if hashState.offset > offset {
// Remove any stored hash state with offsets higher than this one
// as writes to this resumed hasher will make those invalid. This
// is probably okay to skip for now since we don't expect anyone to
// use the API in this way. For that reason, we don't treat an
// an error here as a fatal error, but only log it.
if err := bw.driver.Delete(ctx, hashState.path); err != nil {
logrus.Errorf("unable to delete stale hash state %q: %s", hashState.path, err)
}
}
}
if hashStateMatch.offset == 0 {
// No need to load any state, just reset the hasher.
bw.resumableDigester.Reset()
} else {
storedState, err := bw.driver.GetContent(ctx, hashStateMatch.path)
if err != nil {
return err
}
if err = bw.resumableDigester.Restore(storedState); err != nil {
return err
}
}
// Mind the gap.
if gapLen := offset - int64(bw.resumableDigester.Len()); gapLen > 0 {
// Need to read content from the upload to catch up to the desired offset.
fr, err := newFileReader(ctx, bw.driver, bw.path, bw.size)
if err != nil {
return err
}
if _, err = fr.Seek(int64(bw.resumableDigester.Len()), os.SEEK_SET); err != nil {
return fmt.Errorf("unable to seek to layer reader offset %d: %s", bw.resumableDigester.Len(), err)
}
if _, err := io.CopyN(bw.resumableDigester, fr, gapLen); err != nil {
return err
}
}
return nil
}
func (bw *blobWriter) storeHashState(ctx context.Context) error {
uploadHashStatePath, err := bw.blobStore.pm.path(uploadHashStatePathSpec{
name: bw.blobStore.repository.Name(),
id: bw.id,
alg: bw.resumableDigester.Digest().Algorithm(),
offset: int64(bw.resumableDigester.Len()),
})
if err != nil {
return err
}
hashState, err := bw.resumableDigester.State()
if err != nil {
return err
}
return bw.driver.PutContent(ctx, uploadHashStatePath, hashState)
}
// removeResources should clean up all resources associated with the upload
// instance. An error will be returned if the clean up cannot proceed. If the
// resources are already not present, no error will be returned.
func (bw *blobWriter) removeResources(ctx context.Context) error {
dataPath, err := bw.blobStore.pm.path(uploadDataPathSpec{
name: bw.blobStore.repository.Name(),
id: bw.id,
})
if err != nil {
return err
}
// Resolve and delete the containing directory, which should include any
// upload related files.
dirPath := path.Dir(dataPath)
if err := bw.blobStore.driver.Delete(ctx, dirPath); err != nil {
switch err := err.(type) {
case storagedriver.PathNotFoundError:
break // already gone!
default:
// This should be uncommon enough such that returning an error
// should be okay. At this point, the upload should be mostly
// complete, but perhaps the backend became unaccessible.
context.GetLogger(ctx).Errorf("unable to delete layer upload resources %q: %v", dirPath, err)
return err
}
}
return nil
}

6
docs/storage/blobwriter_nonresumable.go Normal file
View file

@ -0,0 +1,6 @@
// +build noresumabledigest
package storage
func (bw *blobWriter) setupResumableDigester() {
}

9
docs/storage/blobwriter_resumable.go Normal file
View file

@ -0,0 +1,9 @@
// +build !noresumabledigest
package storage
import "github.com/docker/distribution/digest"
func (bw *blobWriter) setupResumableDigester() {
bw.resumableDigester = digest.NewCanonicalResumableDigester()
}

94
docs/storage/cache/cache.go vendored
View file

@ -1,98 +1,38 @@
// Package cache provides facilities to speed up access to the storage
// backend. Typically cache implementations deal with internal implementation
// details at the backend level, rather than generalized caches for
// distribution related interfaces. In other words, unless the cache is
// specific to the storage package, it belongs in another package.
// backend.
package cache
import (
"fmt"
"github.com/docker/distribution"
"github.com/docker/distribution/digest"
"golang.org/x/net/context"
)
// ErrNotFound is returned when a meta item is not found.
var ErrNotFound = fmt.Errorf("not found")
// BlobDescriptorCacheProvider provides repository scoped
// BlobDescriptorService cache instances and a global descriptor cache.
type BlobDescriptorCacheProvider interface {
distribution.BlobDescriptorService
// LayerMeta describes the backend location and length of layer data.
type LayerMeta struct {
Path string
Length int64
RepositoryScoped(repo string) (distribution.BlobDescriptorService, error)
}
// LayerInfoCache is a driver-aware cache of layer metadata. Basically, it
// provides a fast cache for checks against repository metadata, avoiding
// round trips to backend storage. Note that this is different from a pure
// layer cache, which would also provide access to backing data, as well. Such
// a cache should be implemented as a middleware, rather than integrated with
// the storage backend.
//
// Note that most implementations rely on the caller to do strict checks on on
// repo and dgst arguments, since these are mostly used behind existing
// implementations.
type LayerInfoCache interface {
// Contains returns true if the repository with name contains the layer.
Contains(ctx context.Context, repo string, dgst digest.Digest) (bool, error)
// Add includes the layer in the given repository cache.
Add(ctx context.Context, repo string, dgst digest.Digest) error
// Meta provides the location of the layer on the backend and its size. Membership of a
// repository should be tested before using the result, if required.
Meta(ctx context.Context, dgst digest.Digest) (LayerMeta, error)
// SetMeta sets the meta data for the given layer.
SetMeta(ctx context.Context, dgst digest.Digest, meta LayerMeta) error
func validateDigest(dgst digest.Digest) error {
return dgst.Validate()
}
// base implements common checks between cache implementations. Note that
// these are not full checks of input, since that should be done by the
// caller.
type base struct {
LayerInfoCache
func validateDescriptor(desc distribution.Descriptor) error {
if err := validateDigest(desc.Digest); err != nil {
return err
}
func (b *base) Contains(ctx context.Context, repo string, dgst digest.Digest) (bool, error) {
if repo == "" {
return false, fmt.Errorf("cache: cannot check for empty repository name")
if desc.Length < 0 {
return fmt.Errorf("cache: invalid length in descriptor: %v < 0", desc.Length)
}
if dgst == "" {
return false, fmt.Errorf("cache: cannot check for empty digests")
if desc.MediaType == "" {
return fmt.Errorf("cache: empty mediatype on descriptor: %v", desc)
}
return b.LayerInfoCache.Contains(ctx, repo, dgst)
}
func (b *base) Add(ctx context.Context, repo string, dgst digest.Digest) error {
if repo == "" {
return fmt.Errorf("cache: cannot add empty repository name")
}
if dgst == "" {
return fmt.Errorf("cache: cannot add empty digest")
}
return b.LayerInfoCache.Add(ctx, repo, dgst)
}
func (b *base) Meta(ctx context.Context, dgst digest.Digest) (LayerMeta, error) {
if dgst == "" {
return LayerMeta{}, fmt.Errorf("cache: cannot get meta for empty digest")
}
return b.LayerInfoCache.Meta(ctx, dgst)
}
func (b *base) SetMeta(ctx context.Context, dgst digest.Digest, meta LayerMeta) error {
if dgst == "" {
return fmt.Errorf("cache: cannot set meta for empty digest")
}
if meta.Path == "" {
return fmt.Errorf("cache: cannot set empty path for meta")
}
return b.LayerInfoCache.SetMeta(ctx, dgst, meta)
return nil
}

159
docs/storage/cache/cache_test.go vendored
View file

@ -3,84 +3,139 @@ package cache
import (
"testing"
"golang.org/x/net/context"
"github.com/docker/distribution"
"github.com/docker/distribution/context"
"github.com/docker/distribution/digest"
)
// checkLayerInfoCache takes a cache implementation through a common set of
// operations. If adding new tests, please add them here so new
// checkBlobDescriptorCache takes a cache implementation through a common set
// of operations. If adding new tests, please add them here so new
// implementations get the benefit.
func checkLayerInfoCache(t *testing.T, lic LayerInfoCache) {
func checkBlobDescriptorCache(t *testing.T, provider BlobDescriptorCacheProvider) {
ctx := context.Background()
exists, err := lic.Contains(ctx, "", "fake:abc")
checkBlobDescriptorCacheEmptyRepository(t, ctx, provider)
checkBlobDescriptorCacheSetAndRead(t, ctx, provider)
}
func checkBlobDescriptorCacheEmptyRepository(t *testing.T, ctx context.Context, provider BlobDescriptorCacheProvider) {
if _, err := provider.Stat(ctx, "sha384:abc"); err != distribution.ErrBlobUnknown {
t.Fatalf("expected unknown blob error with empty store: %v", err)
}
cache, err := provider.RepositoryScoped("")
if err == nil {
t.Fatalf("expected error checking for cache item with empty repo")
t.Fatalf("expected an error when asking for invalid repo")
}
exists, err = lic.Contains(ctx, "foo/bar", "")
if err == nil {
t.Fatalf("expected error checking for cache item with empty digest")
}
exists, err = lic.Contains(ctx, "foo/bar", "fake:abc")
cache, err = provider.RepositoryScoped("foo/bar")
if err != nil {
t.Fatalf("unexpected error checking for cache item: %v", err)
t.Fatalf("unexpected error getting repository: %v", err)
}
if exists {
t.Fatalf("item should not exist")
if err := cache.SetDescriptor(ctx, "", distribution.Descriptor{
Digest: "sha384:abc",
Length: 10,
MediaType: "application/octet-stream"}); err != digest.ErrDigestInvalidFormat {
t.Fatalf("expected error with invalid digest: %v", err)
}
if err := lic.Add(ctx, "", "fake:abc"); err == nil {
t.Fatalf("expected error adding cache item with empty name")
if err := cache.SetDescriptor(ctx, "sha384:abc", distribution.Descriptor{
Digest: "",
Length: 10,
MediaType: "application/octet-stream"}); err == nil {
t.Fatalf("expected error setting value on invalid descriptor")
}
if err := lic.Add(ctx, "foo/bar", ""); err == nil {
t.Fatalf("expected error adding cache item with empty digest")
if _, err := cache.Stat(ctx, ""); err != digest.ErrDigestInvalidFormat {
t.Fatalf("expected error checking for cache item with empty digest: %v", err)
}
if err := lic.Add(ctx, "foo/bar", "fake:abc"); err != nil {
t.Fatalf("unexpected error adding item: %v", err)
if _, err := cache.Stat(ctx, "sha384:abc"); err != distribution.ErrBlobUnknown {
t.Fatalf("expected unknown blob error with empty repo: %v", err)
}
}
exists, err = lic.Contains(ctx, "foo/bar", "fake:abc")
func checkBlobDescriptorCacheSetAndRead(t *testing.T, ctx context.Context, provider BlobDescriptorCacheProvider) {
localDigest := digest.Digest("sha384:abc")
expected := distribution.Descriptor{
Digest: "sha256:abc",
Length: 10,
MediaType: "application/octet-stream"}
cache, err := provider.RepositoryScoped("foo/bar")
if err != nil {
t.Fatalf("unexpected error checking for cache item: %v", err)
t.Fatalf("unexpected error getting scoped cache: %v", err)
}
if !exists {
t.Fatalf("item should exist")
if err := cache.SetDescriptor(ctx, localDigest, expected); err != nil {
t.Fatalf("error setting descriptor: %v", err)
}
_, err = lic.Meta(ctx, "")
if err == nil || err == ErrNotFound {
t.Fatalf("expected error getting meta for cache item with empty digest")
}
_, err = lic.Meta(ctx, "fake:abc")
if err != ErrNotFound {
t.Fatalf("expected unknown layer error getting meta for cache item with empty digest")
}
if err = lic.SetMeta(ctx, "", LayerMeta{}); err == nil {
t.Fatalf("expected error setting meta for cache item with empty digest")
}
if err = lic.SetMeta(ctx, "foo/bar", LayerMeta{}); err == nil {
t.Fatalf("expected error setting meta for cache item with empty meta")
}
expected := LayerMeta{Path: "/foo/bar", Length: 20}
if err := lic.SetMeta(ctx, "foo/bar", expected); err != nil {
t.Fatalf("unexpected error setting meta: %v", err)
}
meta, err := lic.Meta(ctx, "foo/bar")
desc, err := cache.Stat(ctx, localDigest)
if err != nil {
t.Fatalf("unexpected error getting meta: %v", err)
t.Fatalf("unexpected error statting fake2:abc: %v", err)
}
if meta != expected {
t.Fatalf("retrieved meta data did not match: %v", err)
if expected != desc {
t.Fatalf("unexpected descriptor: %#v != %#v", expected, desc)
}
// also check that we set the canonical key ("fake:abc")
desc, err = cache.Stat(ctx, localDigest)
if err != nil {
t.Fatalf("descriptor not returned for canonical key: %v", err)
}
if expected != desc {
t.Fatalf("unexpected descriptor: %#v != %#v", expected, desc)
}
// ensure that global gets extra descriptor mapping
desc, err = provider.Stat(ctx, localDigest)
if err != nil {
t.Fatalf("expected blob unknown in global cache: %v, %v", err, desc)
}
if desc != expected {
t.Fatalf("unexpected descriptor: %#v != %#v", expected, desc)
}
// get at it through canonical descriptor
desc, err = provider.Stat(ctx, expected.Digest)
if err != nil {
t.Fatalf("unexpected error checking glboal descriptor: %v", err)
}
if desc != expected {
t.Fatalf("unexpected descriptor: %#v != %#v", expected, desc)
}
// now, we set the repo local mediatype to something else and ensure it
// doesn't get changed in the provider cache.
expected.MediaType = "application/json"
if err := cache.SetDescriptor(ctx, localDigest, expected); err != nil {
t.Fatalf("unexpected error setting descriptor: %v", err)
}
desc, err = cache.Stat(ctx, localDigest)
if err != nil {
t.Fatalf("unexpected error getting descriptor: %v", err)
}
if desc != expected {
t.Fatalf("unexpected descriptor: %#v != %#v", desc, expected)
}
desc, err = provider.Stat(ctx, localDigest)
if err != nil {
t.Fatalf("unexpected error getting global descriptor: %v", err)
}
expected.MediaType = "application/octet-stream" // expect original mediatype in global
if desc != expected {
t.Fatalf("unexpected descriptor: %#v != %#v", desc, expected)
}
}

170
docs/storage/cache/memory.go vendored
View file

@ -1,63 +1,149 @@
package cache
import (
"sync"
"github.com/docker/distribution"
"github.com/docker/distribution/context"
"github.com/docker/distribution/digest"
"golang.org/x/net/context"
"github.com/docker/distribution/registry/api/v2"
)
// inmemoryLayerInfoCache is a map-based implementation of LayerInfoCache.
type inmemoryLayerInfoCache struct {
membership map[string]map[digest.Digest]struct{}
meta map[digest.Digest]LayerMeta
type inMemoryBlobDescriptorCacheProvider struct {
global *mapBlobDescriptorCache
repositories map[string]*mapBlobDescriptorCache
mu sync.RWMutex
}
// NewInMemoryLayerInfoCache provides an implementation of LayerInfoCache that
// stores results in memory.
func NewInMemoryLayerInfoCache() LayerInfoCache {
return &base{&inmemoryLayerInfoCache{
membership: make(map[string]map[digest.Digest]struct{}),
meta: make(map[digest.Digest]LayerMeta),
}}
// NewInMemoryBlobDescriptorCacheProvider returns a new mapped-based cache for
// storing blob descriptor data.
func NewInMemoryBlobDescriptorCacheProvider() BlobDescriptorCacheProvider {
return &inMemoryBlobDescriptorCacheProvider{
global: newMapBlobDescriptorCache(),
repositories: make(map[string]*mapBlobDescriptorCache),
}
}
func (ilic *inmemoryLayerInfoCache) Contains(ctx context.Context, repo string, dgst digest.Digest) (bool, error) {
members, ok := ilic.membership[repo]
func (imbdcp *inMemoryBlobDescriptorCacheProvider) RepositoryScoped(repo string) (distribution.BlobDescriptorService, error) {
if err := v2.ValidateRespositoryName(repo); err != nil {
return nil, err
}
imbdcp.mu.RLock()
defer imbdcp.mu.RUnlock()
return &repositoryScopedInMemoryBlobDescriptorCache{
repo: repo,
parent: imbdcp,
repository: imbdcp.repositories[repo],
}, nil
}
func (imbdcp *inMemoryBlobDescriptorCacheProvider) Stat(ctx context.Context, dgst digest.Digest) (distribution.Descriptor, error) {
return imbdcp.global.Stat(ctx, dgst)
}
func (imbdcp *inMemoryBlobDescriptorCacheProvider) SetDescriptor(ctx context.Context, dgst digest.Digest, desc distribution.Descriptor) error {
_, err := imbdcp.Stat(ctx, dgst)
if err == distribution.ErrBlobUnknown {
if dgst.Algorithm() != desc.Digest.Algorithm() && dgst != desc.Digest {
// if the digests differ, set the other canonical mapping
if err := imbdcp.global.SetDescriptor(ctx, desc.Digest, desc); err != nil {
return err
}
}
// unknown, just set it
return imbdcp.global.SetDescriptor(ctx, dgst, desc)
}
// we already know it, do nothing
return err
}
// repositoryScopedInMemoryBlobDescriptorCache provides the request scoped
// repository cache. Instances are not thread-safe but the delegated
// operations are.
type repositoryScopedInMemoryBlobDescriptorCache struct {
repo string
parent *inMemoryBlobDescriptorCacheProvider // allows lazy allocation of repo's map
repository *mapBlobDescriptorCache
}
func (rsimbdcp *repositoryScopedInMemoryBlobDescriptorCache) Stat(ctx context.Context, dgst digest.Digest) (distribution.Descriptor, error) {
if rsimbdcp.repository == nil {
return distribution.Descriptor{}, distribution.ErrBlobUnknown
}
return rsimbdcp.repository.Stat(ctx, dgst)
}
func (rsimbdcp *repositoryScopedInMemoryBlobDescriptorCache) SetDescriptor(ctx context.Context, dgst digest.Digest, desc distribution.Descriptor) error {
if rsimbdcp.repository == nil {
// allocate map since we are setting it now.
rsimbdcp.parent.mu.Lock()
var ok bool
// have to read back value since we may have allocated elsewhere.
rsimbdcp.repository, ok = rsimbdcp.parent.repositories[rsimbdcp.repo]
if !ok {
return false, nil
rsimbdcp.repository = newMapBlobDescriptorCache()
rsimbdcp.parent.repositories[rsimbdcp.repo] = rsimbdcp.repository
}
_, ok = members[dgst]
return ok, nil
rsimbdcp.parent.mu.Unlock()
}
// Add adds the layer to the redis repository blob set.
func (ilic *inmemoryLayerInfoCache) Add(ctx context.Context, repo string, dgst digest.Digest) error {
members, ok := ilic.membership[repo]
if err := rsimbdcp.repository.SetDescriptor(ctx, dgst, desc); err != nil {
return err
}
return rsimbdcp.parent.SetDescriptor(ctx, dgst, desc)
}
// mapBlobDescriptorCache provides a simple map-based implementation of the
// descriptor cache.
type mapBlobDescriptorCache struct {
descriptors map[digest.Digest]distribution.Descriptor
mu sync.RWMutex
}
var _ distribution.BlobDescriptorService = &mapBlobDescriptorCache{}
func newMapBlobDescriptorCache() *mapBlobDescriptorCache {
return &mapBlobDescriptorCache{
descriptors: make(map[digest.Digest]distribution.Descriptor),
}
}
func (mbdc *mapBlobDescriptorCache) Stat(ctx context.Context, dgst digest.Digest) (distribution.Descriptor, error) {
if err := validateDigest(dgst); err != nil {
return distribution.Descriptor{}, err
}
mbdc.mu.RLock()
defer mbdc.mu.RUnlock()
desc, ok := mbdc.descriptors[dgst]
if !ok {
members = make(map[digest.Digest]struct{})
ilic.membership[repo] = members
return distribution.Descriptor{}, distribution.ErrBlobUnknown
}
members[dgst] = struct{}{}
return desc, nil
}
return nil
}
// Meta retrieves the layer meta data from the redis hash, returning
// ErrUnknownLayer if not found.
func (ilic *inmemoryLayerInfoCache) Meta(ctx context.Context, dgst digest.Digest) (LayerMeta, error) {
meta, ok := ilic.meta[dgst]
if !ok {
return LayerMeta{}, ErrNotFound
}
return meta, nil
}
// SetMeta sets the meta data for the given digest using a redis hash. A hash
// is used here since we may store unrelated fields about a layer in the
// future.
func (ilic *inmemoryLayerInfoCache) SetMeta(ctx context.Context, dgst digest.Digest, meta LayerMeta) error {
ilic.meta[dgst] = meta
func (mbdc *mapBlobDescriptorCache) SetDescriptor(ctx context.Context, dgst digest.Digest, desc distribution.Descriptor) error {
if err := validateDigest(dgst); err != nil {
return err
}
if err := validateDescriptor(desc); err != nil {
return err
}
mbdc.mu.Lock()
defer mbdc.mu.Unlock()
mbdc.descriptors[dgst] = desc
return nil
}

6
docs/storage/cache/memory_test.go vendored
View file

@ -2,8 +2,8 @@ package cache
import "testing"
// TestInMemoryLayerInfoCache checks the in memory implementation is working
// TestInMemoryBlobInfoCache checks the in memory implementation is working
// correctly.
func TestInMemoryLayerInfoCache(t *testing.T) {
checkLayerInfoCache(t, NewInMemoryLayerInfoCache())
func TestInMemoryBlobInfoCache(t *testing.T) {
checkBlobDescriptorCache(t, NewInMemoryBlobDescriptorCacheProvider())
}

242
docs/storage/cache/redis.go vendored
View file

@ -1,20 +1,28 @@
package cache
import (
ctxu "github.com/docker/distribution/context"
"fmt"
"github.com/docker/distribution/registry/api/v2"
"github.com/docker/distribution"
"github.com/docker/distribution/context"
"github.com/docker/distribution/digest"
"github.com/garyburd/redigo/redis"
"golang.org/x/net/context"
)
// redisLayerInfoCache provides an implementation of storage.LayerInfoCache
// based on redis. Layer info is stored in two parts. The first provide fast
// access to repository membership through a redis set for each repo. The
// second is a redis hash keyed by the digest of the layer, providing path and
// length information. Note that there is no implied relationship between
// these two caches. The layer may exist in one, both or none and the code
// must be written this way.
type redisLayerInfoCache struct {
// redisBlobStatService provides an implementation of
// BlobDescriptorCacheProvider based on redis. Blob descritors are stored in
// two parts. The first provide fast access to repository membership through a
// redis set for each repo. The second is a redis hash keyed by the digest of
// the layer, providing path, length and mediatype information. There is also
// a per-repository redis hash of the blob descriptor, allowing override of
// data. This is currently used to override the mediatype on a per-repository
// basis.
//
// Note that there is no implied relationship between these two caches. The
// layer may exist in one, both or none and the code must be written this way.
type redisBlobDescriptorService struct {
pool *redis.Pool
// TODO(stevvooe): We use a pool because we don't have great control over
@ -23,76 +31,194 @@ type redisLayerInfoCache struct {
// request objects, we can change this to a connection.
}
// NewRedisLayerInfoCache returns a new redis-based LayerInfoCache using the
// provided redis connection pool.
func NewRedisLayerInfoCache(pool *redis.Pool) LayerInfoCache {
return &base{&redisLayerInfoCache{
var _ BlobDescriptorCacheProvider = &redisBlobDescriptorService{}
// NewRedisBlobDescriptorCacheProvider returns a new redis-based
// BlobDescriptorCacheProvider using the provided redis connection pool.
func NewRedisBlobDescriptorCacheProvider(pool *redis.Pool) BlobDescriptorCacheProvider {
return &redisBlobDescriptorService{
pool: pool,
}}
}
}
// Contains does a membership check on the repository blob set in redis. This
// is used as an access check before looking up global path information. If
// false is returned, the caller should still check the backend to if it
// exists elsewhere.
func (rlic *redisLayerInfoCache) Contains(ctx context.Context, repo string, dgst digest.Digest) (bool, error) {
conn := rlic.pool.Get()
defer conn.Close()
ctxu.GetLogger(ctx).Debugf("(*redisLayerInfoCache).Contains(%q, %q)", repo, dgst)
return redis.Bool(conn.Do("SISMEMBER", rlic.repositoryBlobSetKey(repo), dgst))
// RepositoryScoped returns the scoped cache.
func (rbds *redisBlobDescriptorService) RepositoryScoped(repo string) (distribution.BlobDescriptorService, error) {
if err := v2.ValidateRespositoryName(repo); err != nil {
return nil, err
}
// Add adds the layer to the redis repository blob set.
func (rlic *redisLayerInfoCache) Add(ctx context.Context, repo string, dgst digest.Digest) error {
conn := rlic.pool.Get()
defer conn.Close()
ctxu.GetLogger(ctx).Debugf("(*redisLayerInfoCache).Add(%q, %q)", repo, dgst)
_, err := conn.Do("SADD", rlic.repositoryBlobSetKey(repo), dgst)
return err
return &repositoryScopedRedisBlobDescriptorService{
repo: repo,
upstream: rbds,
}, nil
}
// Meta retrieves the layer meta data from the redis hash, returning
// ErrUnknownLayer if not found.
func (rlic *redisLayerInfoCache) Meta(ctx context.Context, dgst digest.Digest) (LayerMeta, error) {
conn := rlic.pool.Get()
// Stat retrieves the descriptor data from the redis hash entry.
func (rbds *redisBlobDescriptorService) Stat(ctx context.Context, dgst digest.Digest) (distribution.Descriptor, error) {
if err := validateDigest(dgst); err != nil {
return distribution.Descriptor{}, err
}
conn := rbds.pool.Get()
defer conn.Close()
reply, err := redis.Values(conn.Do("HMGET", rlic.blobMetaHashKey(dgst), "path", "length"))
return rbds.stat(ctx, conn, dgst)
}
// stat provides an internal stat call that takes a connection parameter. This
// allows some internal management of the connection scope.
func (rbds *redisBlobDescriptorService) stat(ctx context.Context, conn redis.Conn, dgst digest.Digest) (distribution.Descriptor, error) {
reply, err := redis.Values(conn.Do("HMGET", rbds.blobDescriptorHashKey(dgst), "digest", "length", "mediatype"))
if err != nil {
return LayerMeta{}, err
return distribution.Descriptor{}, err
}
if len(reply) < 2 || reply[0] == nil || reply[1] == nil {
return LayerMeta{}, ErrNotFound
if len(reply) < 2 || reply[0] == nil || reply[1] == nil { // don't care if mediatype is nil
return distribution.Descriptor{}, distribution.ErrBlobUnknown
}
var meta LayerMeta
if _, err := redis.Scan(reply, &meta.Path, &meta.Length); err != nil {
return LayerMeta{}, err
var desc distribution.Descriptor
if _, err := redis.Scan(reply, &desc.Digest, &desc.Length, &desc.MediaType); err != nil {
return distribution.Descriptor{}, err
}
return meta, nil
return desc, nil
}
// SetMeta sets the meta data for the given digest using a redis hash. A hash
// is used here since we may store unrelated fields about a layer in the
// future.
func (rlic *redisLayerInfoCache) SetMeta(ctx context.Context, dgst digest.Digest, meta LayerMeta) error {
conn := rlic.pool.Get()
defer conn.Close()
_, err := conn.Do("HMSET", rlic.blobMetaHashKey(dgst), "path", meta.Path, "length", meta.Length)
// SetDescriptor sets the descriptor data for the given digest using a redis
// hash. A hash is used here since we may store unrelated fields about a layer
// in the future.
func (rbds *redisBlobDescriptorService) SetDescriptor(ctx context.Context, dgst digest.Digest, desc distribution.Descriptor) error {
if err := validateDigest(dgst); err != nil {
return err
}
// repositoryBlobSetKey returns the key for the blob set in the cache.
func (rlic *redisLayerInfoCache) repositoryBlobSetKey(repo string) string {
return "repository::" + repo + "::blobs"
if err := validateDescriptor(desc); err != nil {
return err
}
// blobMetaHashKey returns the cache key for immutable blob meta data.
func (rlic *redisLayerInfoCache) blobMetaHashKey(dgst digest.Digest) string {
conn := rbds.pool.Get()
defer conn.Close()
return rbds.setDescriptor(ctx, conn, dgst, desc)
}
func (rbds *redisBlobDescriptorService) setDescriptor(ctx context.Context, conn redis.Conn, dgst digest.Digest, desc distribution.Descriptor) error {
if _, err := conn.Do("HMSET", rbds.blobDescriptorHashKey(dgst),
"digest", desc.Digest,
"length", desc.Length); err != nil {
return err
}
// Only set mediatype if not already set.
if _, err := conn.Do("HSETNX", rbds.blobDescriptorHashKey(dgst),
"mediatype", desc.MediaType); err != nil {
return err
}
return nil
}
func (rbds *redisBlobDescriptorService) blobDescriptorHashKey(dgst digest.Digest) string {
return "blobs::" + dgst.String()
}
type repositoryScopedRedisBlobDescriptorService struct {
repo string
upstream *redisBlobDescriptorService
}
var _ distribution.BlobDescriptorService = &repositoryScopedRedisBlobDescriptorService{}
// Stat ensures that the digest is a member of the specified repository and
// forwards the descriptor request to the global blob store. If the media type
// differs for the repository, we override it.
func (rsrbds *repositoryScopedRedisBlobDescriptorService) Stat(ctx context.Context, dgst digest.Digest) (distribution.Descriptor, error) {
if err := validateDigest(dgst); err != nil {
return distribution.Descriptor{}, err
}
conn := rsrbds.upstream.pool.Get()
defer conn.Close()
// Check membership to repository first
member, err := redis.Bool(conn.Do("SISMEMBER", rsrbds.repositoryBlobSetKey(rsrbds.repo), dgst))
if err != nil {
return distribution.Descriptor{}, err
}
if !member {
return distribution.Descriptor{}, distribution.ErrBlobUnknown
}
upstream, err := rsrbds.upstream.stat(ctx, conn, dgst)
if err != nil {
return distribution.Descriptor{}, err
}
// We allow a per repository mediatype, let's look it up here.
mediatype, err := redis.String(conn.Do("HGET", rsrbds.blobDescriptorHashKey(dgst), "mediatype"))
if err != nil {
return distribution.Descriptor{}, err
}
if mediatype != "" {
upstream.MediaType = mediatype
}
return upstream, nil
}
func (rsrbds *repositoryScopedRedisBlobDescriptorService) SetDescriptor(ctx context.Context, dgst digest.Digest, desc distribution.Descriptor) error {
if err := validateDigest(dgst); err != nil {
return err
}
if err := validateDescriptor(desc); err != nil {
return err
}
if dgst != desc.Digest {
if dgst.Algorithm() == desc.Digest.Algorithm() {
return fmt.Errorf("redis cache: digest for descriptors differ but algorthim does not: %q != %q", dgst, desc.Digest)
}
}
conn := rsrbds.upstream.pool.Get()
defer conn.Close()
return rsrbds.setDescriptor(ctx, conn, dgst, desc)
}
func (rsrbds *repositoryScopedRedisBlobDescriptorService) setDescriptor(ctx context.Context, conn redis.Conn, dgst digest.Digest, desc distribution.Descriptor) error {
if _, err := conn.Do("SADD", rsrbds.repositoryBlobSetKey(rsrbds.repo), dgst); err != nil {
return err
}
if err := rsrbds.upstream.setDescriptor(ctx, conn, dgst, desc); err != nil {
return err
}
// Override repository mediatype.
if _, err := conn.Do("HSET", rsrbds.blobDescriptorHashKey(dgst), "mediatype", desc.MediaType); err != nil {
return err
}
// Also set the values for the primary descriptor, if they differ by
// algorithm (ie sha256 vs tarsum).
if desc.Digest != "" && dgst != desc.Digest && dgst.Algorithm() != desc.Digest.Algorithm() {
if err := rsrbds.setDescriptor(ctx, conn, desc.Digest, desc); err != nil {
return err
}
}
return nil
}
func (rsrbds *repositoryScopedRedisBlobDescriptorService) blobDescriptorHashKey(dgst digest.Digest) string {
return "repository::" + rsrbds.repo + "::blobs::" + dgst.String()
}
func (rsrbds *repositoryScopedRedisBlobDescriptorService) repositoryBlobSetKey(repo string) string {
return "repository::" + rsrbds.repo + "::blobs"
}

4
docs/storage/cache/redis_test.go vendored
View file

@ -17,7 +17,7 @@ func init() {
// TestRedisLayerInfoCache exercises a live redis instance using the cache
// implementation.
func TestRedisLayerInfoCache(t *testing.T) {
func TestRedisBlobDescriptorCacheProvider(t *testing.T) {
if redisAddr == "" {
// fallback to an environement variable
redisAddr = os.Getenv("TEST_REGISTRY_STORAGE_CACHE_REDIS_ADDR")
@ -46,5 +46,5 @@ func TestRedisLayerInfoCache(t *testing.T) {
t.Fatalf("unexpected error flushing redis db: %v", err)
}
checkLayerInfoCache(t, NewRedisLayerInfoCache(pool))
checkBlobDescriptorCache(t, NewRedisBlobDescriptorCacheProvider(pool))
}

84
docs/storage/cachedblobdescriptorstore.go Normal file
View file

@ -0,0 +1,84 @@
package storage
import (
"expvar"
"sync/atomic"
"github.com/docker/distribution/context"
"github.com/docker/distribution/digest"
"github.com/docker/distribution"
)
type cachedBlobStatter struct {
cache distribution.BlobDescriptorService
backend distribution.BlobStatter
}
func (cbds *cachedBlobStatter) Stat(ctx context.Context, dgst digest.Digest) (distribution.Descriptor, error) {
atomic.AddUint64(&blobStatterCacheMetrics.Stat.Requests, 1)
desc, err := cbds.cache.Stat(ctx, dgst)
if err != nil {
if err != distribution.ErrBlobUnknown {
context.GetLogger(ctx).Errorf("error retrieving descriptor from cache: %v", err)
}
goto fallback
}
atomic.AddUint64(&blobStatterCacheMetrics.Stat.Hits, 1)
return desc, nil
fallback:
atomic.AddUint64(&blobStatterCacheMetrics.Stat.Misses, 1)
desc, err = cbds.backend.Stat(ctx, dgst)
if err != nil {
return desc, err
}
if err := cbds.cache.SetDescriptor(ctx, dgst, desc); err != nil {
context.GetLogger(ctx).Errorf("error adding descriptor %v to cache: %v", desc.Digest, err)
}
return desc, err
}
// blobStatterCacheMetrics keeps track of cache metrics for blob descriptor
// cache requests. Note this is kept globally and made available via expvar.
// For more detailed metrics, its recommend to instrument a particular cache
// implementation.
var blobStatterCacheMetrics struct {
// Stat tracks calls to the caches.
Stat struct {
Requests uint64
Hits uint64
Misses uint64
}
}
func init() {
registry := expvar.Get("registry")
if registry == nil {
registry = expvar.NewMap("registry")
}
cache := registry.(*expvar.Map).Get("cache")
if cache == nil {
cache = &expvar.Map{}
cache.(*expvar.Map).Init()
registry.(*expvar.Map).Set("cache", cache)
}
storage := cache.(*expvar.Map).Get("storage")
if storage == nil {
storage = &expvar.Map{}
storage.(*expvar.Map).Init()
cache.(*expvar.Map).Set("storage", storage)
}
storage.(*expvar.Map).Set("blobdescriptor", expvar.Func(func() interface{} {
// no need for synchronous access: the increments are atomic and
// during reading, we don't care if the data is up to date. The
// numbers will always *eventually* be reported correctly.
return blobStatterCacheMetrics
}))
}

49
docs/storage/filereader.go
View file

@ -7,7 +7,6 @@ import (
"io"
"io/ioutil"
"os"
"time"
"github.com/docker/distribution/context"
storagedriver "github.com/docker/distribution/registry/storage/driver"
@ -31,7 +30,6 @@ type fileReader struct {
// identifying fields
path string
size int64 // size is the total size, must be set.
modtime time.Time // TODO(stevvooe): This is not needed anymore.
// mutable fields
rc io.ReadCloser // remote read closer
@ -40,41 +38,17 @@ type fileReader struct {
err error // terminal error, if set, reader is closed
}
// newFileReader initializes a file reader for the remote file. The read takes
// on the offset and size at the time the reader is created. If the underlying
// file changes, one must create a new fileReader.
func newFileReader(ctx context.Context, driver storagedriver.StorageDriver, path string) (*fileReader, error) {
rd := &fileReader{
// newFileReader initializes a file reader for the remote file. The reader
// takes on the size and path that must be determined externally with a stat
// call. The reader operates optimistically, assuming that the file is already
// there.
func newFileReader(ctx context.Context, driver storagedriver.StorageDriver, path string, size int64) (*fileReader, error) {
return &fileReader{
ctx: ctx,
driver: driver,
path: path,
ctx: ctx,
}
// Grab the size of the layer file, ensuring existence.
if fi, err := driver.Stat(ctx, path); err != nil {
switch err := err.(type) {
case storagedriver.PathNotFoundError:
// NOTE(stevvooe): We really don't care if the file is not
// actually present for the reader. If the caller needs to know
// whether or not the file exists, they should issue a stat call
// on the path. There is still no guarantee, since the file may be
// gone by the time the reader is created. The only correct
// behavior is to return a reader that immediately returns EOF.
default:
// Any other error we want propagated up the stack.
return nil, err
}
} else {
if fi.IsDir() {
return nil, fmt.Errorf("cannot read a directory")
}
// Fill in file information
rd.size = fi.Size()
rd.modtime = fi.ModTime()
}
return rd, nil
size: size,
}, nil
}
func (fr *fileReader) Read(p []byte) (n int, err error) {
@ -162,11 +136,6 @@ func (fr *fileReader) reader() (io.Reader, error) {
fr.rc = rc
if fr.brd == nil {
// TODO(stevvooe): Set an optimal buffer size here. We'll have to
// understand the latency characteristics of the underlying network to
// set this correctly, so we may want to leave it to the driver. For
// out of process drivers, we'll have to optimize this buffer size for
// local communication.
fr.brd = bufio.NewReaderSize(fr.rc, fileReaderBufferSize)
} else {
fr.brd.Reset(fr.rc)

6
docs/storage/filereader_test.go
View file

@ -37,7 +37,7 @@ func TestSimpleRead(t *testing.T) {
t.Fatalf("error putting patterned content: %v", err)
}
fr, err := newFileReader(ctx, driver, path)
fr, err := newFileReader(ctx, driver, path, int64(len(content)))
if err != nil {
t.Fatalf("error allocating file reader: %v", err)
}
@ -66,7 +66,7 @@ func TestFileReaderSeek(t *testing.T) {
t.Fatalf("error putting patterned content: %v", err)
}
fr, err := newFileReader(ctx, driver, path)
fr, err := newFileReader(ctx, driver, path, int64(len(content)))
if err != nil {
t.Fatalf("unexpected error creating file reader: %v", err)
@ -162,7 +162,7 @@ func TestFileReaderSeek(t *testing.T) {
// read method, with an io.EOF error.
func TestFileReaderNonExistentFile(t *testing.T) {
driver := inmemory.New()
fr, err := newFileReader(context.Background(), driver, "/doesnotexist")
fr, err := newFileReader(context.Background(), driver, "/doesnotexist", 10)
if err != nil {
t.Fatalf("unexpected error initializing reader: %v", err)
}

60
docs/storage/filewriter.go
View file

@ -39,7 +39,6 @@ type bufferedFileWriter struct {
// filewriter should implement.
type fileWriterInterface interface {
io.WriteSeeker
io.WriterAt
io.ReaderFrom
io.Closer
}
@ -110,21 +109,31 @@ func (bfw *bufferedFileWriter) Flush() error {
// Write writes the buffer p at the current write offset.
func (fw *fileWriter) Write(p []byte) (n int, err error) {
nn, err := fw.readFromAt(bytes.NewReader(p), -1)
return int(nn), err
}
// WriteAt writes p at the specified offset. The underlying offset does not
// change.
func (fw *fileWriter) WriteAt(p []byte, offset int64) (n int, err error) {
nn, err := fw.readFromAt(bytes.NewReader(p), offset)
nn, err := fw.ReadFrom(bytes.NewReader(p))
return int(nn), err
}
// ReadFrom reads reader r until io.EOF writing the contents at the current
// offset.
func (fw *fileWriter) ReadFrom(r io.Reader) (n int64, err error) {
return fw.readFromAt(r, -1)
if fw.err != nil {
return 0, fw.err
}
nn, err := fw.driver.WriteStream(fw.ctx, fw.path, fw.offset, r)
// We should forward the offset, whether or not there was an error.
// Basically, we keep the filewriter in sync with the reader's head. If an
// error is encountered, the whole thing should be retried but we proceed
// from an expected offset, even if the data didn't make it to the
// backend.
fw.offset += nn
if fw.offset > fw.size {
fw.size = fw.offset
}
return nn, err
}
// Seek moves the write position do the requested offest based on the whence
@ -169,34 +178,3 @@ func (fw *fileWriter) Close() error {
return nil
}
// readFromAt writes to fw from r at the specified offset. If offset is less
// than zero, the value of fw.offset is used and updated after the operation.
func (fw *fileWriter) readFromAt(r io.Reader, offset int64) (n int64, err error) {
if fw.err != nil {
return 0, fw.err
}
var updateOffset bool
if offset < 0 {
offset = fw.offset
updateOffset = true
}
nn, err := fw.driver.WriteStream(fw.ctx, fw.path, offset, r)
if updateOffset {
// We should forward the offset, whether or not there was an error.
// Basically, we keep the filewriter in sync with the reader's head. If an
// error is encountered, the whole thing should be retried but we proceed
// from an expected offset, even if the data didn't make it to the
// backend.
fw.offset += nn
if fw.offset > fw.size {
fw.size = fw.offset
}
}
return nn, err
}

24
docs/storage/filewriter_test.go
View file

@ -51,7 +51,7 @@ func TestSimpleWrite(t *testing.T) {
t.Fatalf("unexpected write length: %d != %d", n, len(content))
}
fr, err := newFileReader(ctx, driver, path)
fr, err := newFileReader(ctx, driver, path, int64(len(content)))
if err != nil {
t.Fatalf("unexpected error creating fileReader: %v", err)
}
@ -78,23 +78,23 @@ func TestSimpleWrite(t *testing.T) {
t.Fatalf("write did not advance offset: %d != %d", end, len(content))
}
// Double the content, but use the WriteAt method
// Double the content
doubled := append(content, content...)
doubledgst, err := digest.FromReader(bytes.NewReader(doubled))
if err != nil {
t.Fatalf("unexpected error digesting doubled content: %v", err)
}
n, err = fw.WriteAt(content, end)
nn, err := fw.ReadFrom(bytes.NewReader(content))
if err != nil {
t.Fatalf("unexpected error writing content at %d: %v", end, err)
t.Fatalf("unexpected error doubling content: %v", err)
}
if n != len(content) {
if nn != int64(len(content)) {
t.Fatalf("writeat was short: %d != %d", n, len(content))
}
fr, err = newFileReader(ctx, driver, path)
fr, err = newFileReader(ctx, driver, path, int64(len(doubled)))
if err != nil {
t.Fatalf("unexpected error creating fileReader: %v", err)
}
@ -111,20 +111,20 @@ func TestSimpleWrite(t *testing.T) {
t.Fatalf("unable to verify write data")
}
// Check that WriteAt didn't update the offset.
// Check that Write updated the offset.
end, err = fw.Seek(0, os.SEEK_END)
if err != nil {
t.Fatalf("unexpected error seeking: %v", err)
}
if end != int64(len(content)) {
t.Fatalf("write did not advance offset: %d != %d", end, len(content))
if end != int64(len(doubled)) {
t.Fatalf("write did not advance offset: %d != %d", end, len(doubled))
}
// Now, we copy from one path to another, running the data through the
// fileReader to fileWriter, rather than the driver.Move command to ensure
// everything is working correctly.
fr, err = newFileReader(ctx, driver, path)
fr, err = newFileReader(ctx, driver, path, int64(len(doubled)))
if err != nil {
t.Fatalf("unexpected error creating fileReader: %v", err)
}
@ -136,7 +136,7 @@ func TestSimpleWrite(t *testing.T) {
}
defer fw.Close()
nn, err := io.Copy(fw, fr)
nn, err = io.Copy(fw, fr)
if err != nil {
t.Fatalf("unexpected error copying data: %v", err)
}
@ -145,7 +145,7 @@ func TestSimpleWrite(t *testing.T) {
t.Fatalf("unexpected copy length: %d != %d", nn, len(doubled))
}
fr, err = newFileReader(ctx, driver, "/copied")
fr, err = newFileReader(ctx, driver, "/copied", int64(len(doubled)))
if err != nil {
t.Fatalf("unexpected error creating fileReader: %v", err)
}

202
docs/storage/layercache.go
View file

@ -1,202 +0,0 @@
package storage
import (
"expvar"
"sync/atomic"
"time"
"github.com/docker/distribution"
ctxu "github.com/docker/distribution/context"
"github.com/docker/distribution/digest"
"github.com/docker/distribution/registry/storage/cache"
"github.com/docker/distribution/registry/storage/driver"
"golang.org/x/net/context"
)
// cachedLayerService implements the layer service with path-aware caching,
// using a LayerInfoCache interface.
type cachedLayerService struct {
distribution.LayerService // upstream layer service
repository distribution.Repository
ctx context.Context
driver driver.StorageDriver
*blobStore // global blob store
cache cache.LayerInfoCache
}
// Exists checks for existence of the digest in the cache, immediately
// returning if it exists for the repository. If not, the upstream is checked.
// When a positive result is found, it is written into the cache.
func (lc *cachedLayerService) Exists(dgst digest.Digest) (bool, error) {
ctxu.GetLogger(lc.ctx).Debugf("(*cachedLayerService).Exists(%q)", dgst)
now := time.Now()
defer func() {
// TODO(stevvooe): Replace this with a decent context-based metrics solution
ctxu.GetLoggerWithField(lc.ctx, "blob.exists.duration", time.Since(now)).
Infof("(*cachedLayerService).Exists(%q)", dgst)
}()
atomic.AddUint64(&layerInfoCacheMetrics.Exists.Requests, 1)
available, err := lc.cache.Contains(lc.ctx, lc.repository.Name(), dgst)
if err != nil {
ctxu.GetLogger(lc.ctx).Errorf("error checking availability of %v@%v: %v", lc.repository.Name(), dgst, err)
goto fallback
}
if available {
atomic.AddUint64(&layerInfoCacheMetrics.Exists.Hits, 1)
return true, nil
}
fallback:
atomic.AddUint64(&layerInfoCacheMetrics.Exists.Misses, 1)
exists, err := lc.LayerService.Exists(dgst)
if err != nil {
return exists, err
}
if exists {
// we can only cache this if the existence is positive.
if err := lc.cache.Add(lc.ctx, lc.repository.Name(), dgst); err != nil {
ctxu.GetLogger(lc.ctx).Errorf("error adding %v@%v to cache: %v", lc.repository.Name(), dgst, err)
}
}
return exists, err
}
// Fetch checks for the availability of the layer in the repository via the
// cache. If present, the metadata is resolved and the layer is returned. If
// any operation fails, the layer is read directly from the upstream. The
// results are cached, if possible.
func (lc *cachedLayerService) Fetch(dgst digest.Digest) (distribution.Layer, error) {
ctxu.GetLogger(lc.ctx).Debugf("(*layerInfoCache).Fetch(%q)", dgst)
now := time.Now()
defer func() {
ctxu.GetLoggerWithField(lc.ctx, "blob.fetch.duration", time.Since(now)).
Infof("(*layerInfoCache).Fetch(%q)", dgst)
}()
atomic.AddUint64(&layerInfoCacheMetrics.Fetch.Requests, 1)
available, err := lc.cache.Contains(lc.ctx, lc.repository.Name(), dgst)
if err != nil {
ctxu.GetLogger(lc.ctx).Errorf("error checking availability of %v@%v: %v", lc.repository.Name(), dgst, err)
goto fallback
}
if available {
// fast path: get the layer info and return
meta, err := lc.cache.Meta(lc.ctx, dgst)
if err != nil {
ctxu.GetLogger(lc.ctx).Errorf("error fetching %v@%v from cache: %v", lc.repository.Name(), dgst, err)
goto fallback
}
atomic.AddUint64(&layerInfoCacheMetrics.Fetch.Hits, 1)
return newLayerReader(lc.driver, dgst, meta.Path, meta.Length)
}
// NOTE(stevvooe): Unfortunately, the cache here only makes checks for
// existing layers faster. We'd have to provide more careful
// synchronization with the backend to make the missing case as fast.
fallback:
atomic.AddUint64(&layerInfoCacheMetrics.Fetch.Misses, 1)
layer, err := lc.LayerService.Fetch(dgst)
if err != nil {
return nil, err
}
// add the layer to the repository
if err := lc.cache.Add(lc.ctx, lc.repository.Name(), dgst); err != nil {
ctxu.GetLogger(lc.ctx).
Errorf("error caching repository relationship for %v@%v: %v", lc.repository.Name(), dgst, err)
}
// lookup layer path and add it to the cache, if it succeds. Note that we
// still return the layer even if we have trouble caching it.
if path, err := lc.resolveLayerPath(layer); err != nil {
ctxu.GetLogger(lc.ctx).
Errorf("error resolving path while caching %v@%v: %v", lc.repository.Name(), dgst, err)
} else {
// add the layer to the cache once we've resolved the path.
if err := lc.cache.SetMeta(lc.ctx, dgst, cache.LayerMeta{Path: path, Length: layer.Length()}); err != nil {
ctxu.GetLogger(lc.ctx).Errorf("error adding meta for %v@%v to cache: %v", lc.repository.Name(), dgst, err)
}
}
return layer, err
}
// extractLayerInfo pulls the layerInfo from the layer, attempting to get the
// path information from either the concrete object or by resolving the
// primary blob store path.
func (lc *cachedLayerService) resolveLayerPath(layer distribution.Layer) (path string, err error) {
// try and resolve the type and driver, so we don't have to traverse links
switch v := layer.(type) {
case *layerReader:
// only set path if we have same driver instance.
if v.driver == lc.driver {
return v.path, nil
}
}
ctxu.GetLogger(lc.ctx).Warnf("resolving layer path during cache lookup (%v@%v)", lc.repository.Name(), layer.Digest())
// we have to do an expensive stat to resolve the layer location but no
// need to check the link, since we already have layer instance for this
// repository.
bp, err := lc.blobStore.path(layer.Digest())
if err != nil {
return "", err
}
return bp, nil
}
// layerInfoCacheMetrics keeps track of cache metrics for layer info cache
// requests. Note this is kept globally and made available via expvar. For
// more detailed metrics, its recommend to instrument a particular cache
// implementation.
var layerInfoCacheMetrics struct {
// Exists tracks calls to the Exists caches.
Exists struct {
Requests uint64
Hits uint64
Misses uint64
}
// Fetch tracks calls to the fetch caches.
Fetch struct {
Requests uint64
Hits uint64
Misses uint64
}
}
func init() {
registry := expvar.Get("registry")
if registry == nil {
registry = expvar.NewMap("registry")
}
cache := registry.(*expvar.Map).Get("cache")
if cache == nil {
cache = &expvar.Map{}
cache.(*expvar.Map).Init()
registry.(*expvar.Map).Set("cache", cache)
}
storage := cache.(*expvar.Map).Get("storage")
if storage == nil {
storage = &expvar.Map{}
storage.(*expvar.Map).Init()
cache.(*expvar.Map).Set("storage", storage)
}
storage.(*expvar.Map).Set("layerinfo", expvar.Func(func() interface{} {
// no need for synchronous access: the increments are atomic and
// during reading, we don't care if the data is up to date. The
// numbers will always *eventually* be reported correctly.
return layerInfoCacheMetrics
}))
}

104
docs/storage/layerreader.go
View file

@ -1,104 +0,0 @@
package storage
import (
"fmt"
"net/http"
"time"
"github.com/docker/distribution"
"github.com/docker/distribution/digest"
"github.com/docker/distribution/registry/storage/driver"
)
// layerReader implements Layer and provides facilities for reading and
// seeking.
type layerReader struct {
fileReader
digest digest.Digest
}
// newLayerReader returns a new layerReader with the digest, path and length,
// eliding round trips to the storage backend.
func newLayerReader(driver driver.StorageDriver, dgst digest.Digest, path string, length int64) (*layerReader, error) {
fr := &fileReader{
driver: driver,
path: path,
size: length,
}
return &layerReader{
fileReader: *fr,
digest: dgst,
}, nil
}
var _ distribution.Layer = &layerReader{}
func (lr *layerReader) Digest() digest.Digest {
return lr.digest
}
func (lr *layerReader) Length() int64 {
return lr.size
}
func (lr *layerReader) CreatedAt() time.Time {
return lr.modtime
}
// Close the layer. Should be called when the resource is no longer needed.
func (lr *layerReader) Close() error {
return lr.closeWithErr(distribution.ErrLayerClosed)
}
func (lr *layerReader) Handler(r *http.Request) (h http.Handler, err error) {
var handlerFunc http.HandlerFunc
redirectURL, err := lr.fileReader.driver.URLFor(lr.ctx, lr.path, map[string]interface{}{"method": r.Method})
switch err {
case nil:
handlerFunc = func(w http.ResponseWriter, r *http.Request) {
// Redirect to storage URL.
http.Redirect(w, r, redirectURL, http.StatusTemporaryRedirect)
}
case driver.ErrUnsupportedMethod:
handlerFunc = func(w http.ResponseWriter, r *http.Request) {
// Fallback to serving the content directly.
http.ServeContent(w, r, lr.digest.String(), lr.CreatedAt(), lr)
}
default:
// Some unexpected error.
return nil, err
}
return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
// If the registry is serving this content itself, check
// the If-None-Match header and return 304 on match. Redirected
// storage implementations do the same.
if etagMatch(r, lr.digest.String()) {
w.WriteHeader(http.StatusNotModified)
return
}
setCacheHeaders(w, 86400, lr.digest.String())
w.Header().Set("Docker-Content-Digest", lr.digest.String())
handlerFunc.ServeHTTP(w, r)
}), nil
}
func etagMatch(r *http.Request, etag string) bool {
for _, headerVal := range r.Header["If-None-Match"] {
if headerVal == etag {
return true
}
}
return false
}
func setCacheHeaders(w http.ResponseWriter, cacheAge int, etag string) {
w.Header().Set("ETag", etag)
w.Header().Set("Cache-Control", fmt.Sprintf("max-age=%d", cacheAge))
}

178
docs/storage/layerstore.go
View file

@ -1,178 +0,0 @@
package storage
import (
"time"
"code.google.com/p/go-uuid/uuid"
"github.com/docker/distribution"
"github.com/docker/distribution/context"
"github.com/docker/distribution/digest"
"github.com/docker/distribution/manifest"
storagedriver "github.com/docker/distribution/registry/storage/driver"
)
type layerStore struct {
repository *repository
}
func (ls *layerStore) Exists(digest digest.Digest) (bool, error) {
context.GetLogger(ls.repository.ctx).Debug("(*layerStore).Exists")
// Because this implementation just follows blob links, an existence check
// is pretty cheap by starting and closing a fetch.
_, err := ls.Fetch(digest)
if err != nil {
switch err.(type) {
case distribution.ErrUnknownLayer:
return false, nil
}
return false, err
}
return true, nil
}
func (ls *layerStore) Fetch(dgst digest.Digest) (distribution.Layer, error) {
ctx := ls.repository.ctx
context.GetLogger(ctx).Debug("(*layerStore).Fetch")
bp, err := ls.path(dgst)
if err != nil {
return nil, err
}
fr, err := newFileReader(ctx, ls.repository.driver, bp)
if err != nil {
return nil, err
}
return &layerReader{
fileReader: *fr,
digest: dgst,
}, nil
}
// Upload begins a layer upload, returning a handle. If the layer upload
// is already in progress or the layer has already been uploaded, this
// will return an error.
func (ls *layerStore) Upload() (distribution.LayerUpload, error) {
ctx := ls.repository.ctx
context.GetLogger(ctx).Debug("(*layerStore).Upload")
// NOTE(stevvooe): Consider the issues with allowing concurrent upload of
// the same two layers. Should it be disallowed? For now, we allow both
// parties to proceed and the the first one uploads the layer.
uuid := uuid.New()
startedAt := time.Now().UTC()
path, err := ls.repository.pm.path(uploadDataPathSpec{
name: ls.repository.Name(),
uuid: uuid,
})
if err != nil {
return nil, err
}
startedAtPath, err := ls.repository.pm.path(uploadStartedAtPathSpec{
name: ls.repository.Name(),
uuid: uuid,
})
if err != nil {
return nil, err
}
// Write a startedat file for this upload
if err := ls.repository.driver.PutContent(ctx, startedAtPath, []byte(startedAt.Format(time.RFC3339))); err != nil {
return nil, err
}
return ls.newLayerUpload(uuid, path, startedAt)
}
// Resume continues an in progress layer upload, returning the current
// state of the upload.
func (ls *layerStore) Resume(uuid string) (distribution.LayerUpload, error) {
ctx := ls.repository.ctx
context.GetLogger(ctx).Debug("(*layerStore).Resume")
startedAtPath, err := ls.repository.pm.path(uploadStartedAtPathSpec{
name: ls.repository.Name(),
uuid: uuid,
})
if err != nil {
return nil, err
}
startedAtBytes, err := ls.repository.driver.GetContent(ctx, startedAtPath)
if err != nil {
switch err := err.(type) {
case storagedriver.PathNotFoundError:
return nil, distribution.ErrLayerUploadUnknown
default:
return nil, err
}
}
startedAt, err := time.Parse(time.RFC3339, string(startedAtBytes))
if err != nil {
return nil, err
}
path, err := ls.repository.pm.path(uploadDataPathSpec{
name: ls.repository.Name(),
uuid: uuid,
})
if err != nil {
return nil, err
}
return ls.newLayerUpload(uuid, path, startedAt)
}
// newLayerUpload allocates a new upload controller with the given state.
func (ls *layerStore) newLayerUpload(uuid, path string, startedAt time.Time) (distribution.LayerUpload, error) {
fw, err := newFileWriter(ls.repository.ctx, ls.repository.driver, path)
if err != nil {
return nil, err
}
lw := &layerWriter{
layerStore: ls,
uuid: uuid,
startedAt: startedAt,
bufferedFileWriter: *fw,
}
lw.setupResumableDigester()
return lw, nil
}
func (ls *layerStore) path(dgst digest.Digest) (string, error) {
// We must traverse this path through the link to enforce ownership.
layerLinkPath, err := ls.repository.pm.path(layerLinkPathSpec{name: ls.repository.Name(), digest: dgst})
if err != nil {
return "", err
}
blobPath, err := ls.repository.blobStore.resolve(layerLinkPath)
if err != nil {
switch err := err.(type) {
case storagedriver.PathNotFoundError:
return "", distribution.ErrUnknownLayer{
FSLayer: manifest.FSLayer{BlobSum: dgst},
}
default:
return "", err
}
}
return blobPath, nil
}

478
docs/storage/layerwriter.go
View file

@ -1,478 +0,0 @@
package storage
import (
"fmt"
"io"
"os"
"path"
"strconv"
"time"
"github.com/Sirupsen/logrus"
"github.com/docker/distribution"
"github.com/docker/distribution/context"
"github.com/docker/distribution/digest"
storagedriver "github.com/docker/distribution/registry/storage/driver"
)
var _ distribution.LayerUpload = &layerWriter{}
// layerWriter is used to control the various aspects of resumable
// layer upload. It implements the LayerUpload interface.
type layerWriter struct {
layerStore *layerStore
uuid string
startedAt time.Time
resumableDigester digest.ResumableDigester
// implementes io.WriteSeeker, io.ReaderFrom and io.Closer to satisfy
// LayerUpload Interface
bufferedFileWriter
}
var _ distribution.LayerUpload = &layerWriter{}
// UUID returns the identifier for this upload.
func (lw *layerWriter) UUID() string {
return lw.uuid
}
func (lw *layerWriter) StartedAt() time.Time {
return lw.startedAt
}
// Finish marks the upload as completed, returning a valid handle to the
// uploaded layer. The final size and checksum are validated against the
// contents of the uploaded layer. The checksum should be provided in the
// format <algorithm>:<hex digest>.
func (lw *layerWriter) Finish(dgst digest.Digest) (distribution.Layer, error) {
context.GetLogger(lw.layerStore.repository.ctx).Debug("(*layerWriter).Finish")
if err := lw.bufferedFileWriter.Close(); err != nil {
return nil, err
}
var (
canonical digest.Digest
err error
)
// HACK(stevvooe): To deal with s3's lack of consistency, attempt to retry
// validation on failure. Three attempts are made, backing off
// retries*100ms each time.
for retries := 0; ; retries++ {
canonical, err = lw.validateLayer(dgst)
if err == nil {
break
}
context.GetLoggerWithField(lw.layerStore.repository.ctx, "retries", retries).
Errorf("error validating layer: %v", err)
if retries < 3 {
time.Sleep(100 * time.Millisecond * time.Duration(retries+1))
continue
}
return nil, err
}
if err := lw.moveLayer(canonical); err != nil {
// TODO(stevvooe): Cleanup?
return nil, err
}
// Link the layer blob into the repository.
if err := lw.linkLayer(canonical, dgst); err != nil {
return nil, err
}
if err := lw.removeResources(); err != nil {
return nil, err
}
return lw.layerStore.Fetch(canonical)
}
// Cancel the layer upload process.
func (lw *layerWriter) Cancel() error {
context.GetLogger(lw.layerStore.repository.ctx).Debug("(*layerWriter).Cancel")
if err := lw.removeResources(); err != nil {
return err
}
lw.Close()
return nil
}
func (lw *layerWriter) Write(p []byte) (int, error) {
if lw.resumableDigester == nil {
return lw.bufferedFileWriter.Write(p)
}
// Ensure that the current write offset matches how many bytes have been
// written to the digester. If not, we need to update the digest state to
// match the current write position.
if err := lw.resumeHashAt(lw.offset); err != nil {
return 0, err
}
return io.MultiWriter(&lw.bufferedFileWriter, lw.resumableDigester).Write(p)
}
func (lw *layerWriter) ReadFrom(r io.Reader) (n int64, err error) {
if lw.resumableDigester == nil {
return lw.bufferedFileWriter.ReadFrom(r)
}
// Ensure that the current write offset matches how many bytes have been
// written to the digester. If not, we need to update the digest state to
// match the current write position.
if err := lw.resumeHashAt(lw.offset); err != nil {
return 0, err
}
return lw.bufferedFileWriter.ReadFrom(io.TeeReader(r, lw.resumableDigester))
}
func (lw *layerWriter) Close() error {
if lw.err != nil {
return lw.err
}
if lw.resumableDigester != nil {
if err := lw.storeHashState(); err != nil {
return err
}
}
return lw.bufferedFileWriter.Close()
}
type hashStateEntry struct {
offset int64
path string
}
// getStoredHashStates returns a slice of hashStateEntries for this upload.
func (lw *layerWriter) getStoredHashStates() ([]hashStateEntry, error) {
uploadHashStatePathPrefix, err := lw.layerStore.repository.pm.path(uploadHashStatePathSpec{
name: lw.layerStore.repository.Name(),
uuid: lw.uuid,
alg: lw.resumableDigester.Digest().Algorithm(),
list: true,
})
if err != nil {
return nil, err
}
paths, err := lw.driver.List(lw.layerStore.repository.ctx, uploadHashStatePathPrefix)
if err != nil {
if _, ok := err.(storagedriver.PathNotFoundError); !ok {
return nil, err
}
// Treat PathNotFoundError as no entries.
paths = nil
}
hashStateEntries := make([]hashStateEntry, 0, len(paths))
for _, p := range paths {
pathSuffix := path.Base(p)
// The suffix should be the offset.
offset, err := strconv.ParseInt(pathSuffix, 0, 64)
if err != nil {
logrus.Errorf("unable to parse offset from upload state path %q: %s", p, err)
}
hashStateEntries = append(hashStateEntries, hashStateEntry{offset: offset, path: p})
}
return hashStateEntries, nil
}
// resumeHashAt attempts to restore the state of the internal hash function
// by loading the most recent saved hash state less than or equal to the given
// offset. Any unhashed bytes remaining less than the given offset are hashed
// from the content uploaded so far.
func (lw *layerWriter) resumeHashAt(offset int64) error {
if offset < 0 {
return fmt.Errorf("cannot resume hash at negative offset: %d", offset)
}
if offset == int64(lw.resumableDigester.Len()) {
// State of digester is already at the requested offset.
return nil
}
// List hash states from storage backend.
var hashStateMatch hashStateEntry
hashStates, err := lw.getStoredHashStates()
if err != nil {
return fmt.Errorf("unable to get stored hash states with offset %d: %s", offset, err)
}
ctx := lw.layerStore.repository.ctx
// Find the highest stored hashState with offset less than or equal to
// the requested offset.
for _, hashState := range hashStates {
if hashState.offset == offset {
hashStateMatch = hashState
break // Found an exact offset match.
} else if hashState.offset < offset && hashState.offset > hashStateMatch.offset {
// This offset is closer to the requested offset.
hashStateMatch = hashState
} else if hashState.offset > offset {
// Remove any stored hash state with offsets higher than this one
// as writes to this resumed hasher will make those invalid. This
// is probably okay to skip for now since we don't expect anyone to
// use the API in this way. For that reason, we don't treat an
// an error here as a fatal error, but only log it.
if err := lw.driver.Delete(ctx, hashState.path); err != nil {
logrus.Errorf("unable to delete stale hash state %q: %s", hashState.path, err)
}
}
}
if hashStateMatch.offset == 0 {
// No need to load any state, just reset the hasher.
lw.resumableDigester.Reset()
} else {
storedState, err := lw.driver.GetContent(ctx, hashStateMatch.path)
if err != nil {
return err
}
if err = lw.resumableDigester.Restore(storedState); err != nil {
return err
}
}
// Mind the gap.
if gapLen := offset - int64(lw.resumableDigester.Len()); gapLen > 0 {
// Need to read content from the upload to catch up to the desired offset.
fr, err := newFileReader(ctx, lw.driver, lw.path)
if err != nil {
return err
}
if _, err = fr.Seek(int64(lw.resumableDigester.Len()), os.SEEK_SET); err != nil {
return fmt.Errorf("unable to seek to layer reader offset %d: %s", lw.resumableDigester.Len(), err)
}
if _, err := io.CopyN(lw.resumableDigester, fr, gapLen); err != nil {
return err
}
}
return nil
}
func (lw *layerWriter) storeHashState() error {
uploadHashStatePath, err := lw.layerStore.repository.pm.path(uploadHashStatePathSpec{
name: lw.layerStore.repository.Name(),
uuid: lw.uuid,
alg: lw.resumableDigester.Digest().Algorithm(),
offset: int64(lw.resumableDigester.Len()),
})
if err != nil {
return err
}
hashState, err := lw.resumableDigester.State()
if err != nil {
return err
}
return lw.driver.PutContent(lw.layerStore.repository.ctx, uploadHashStatePath, hashState)
}
// validateLayer checks the layer data against the digest, returning an error
// if it does not match. The canonical digest is returned.
func (lw *layerWriter) validateLayer(dgst digest.Digest) (digest.Digest, error) {
var (
verified, fullHash bool
canonical digest.Digest
)
if lw.resumableDigester != nil {
// Restore the hasher state to the end of the upload.
if err := lw.resumeHashAt(lw.size); err != nil {
return "", err
}
canonical = lw.resumableDigester.Digest()
if canonical.Algorithm() == dgst.Algorithm() {
// Common case: client and server prefer the same canonical digest
// algorithm - currently SHA256.
verified = dgst == canonical
} else {
// The client wants to use a different digest algorithm. They'll just
// have to be patient and wait for us to download and re-hash the
// uploaded content using that digest algorithm.
fullHash = true
}
} else {
// Not using resumable digests, so we need to hash the entire layer.
fullHash = true
}
if fullHash {
digester := digest.NewCanonicalDigester()
digestVerifier, err := digest.NewDigestVerifier(dgst)
if err != nil {
return "", err
}
// Read the file from the backend driver and validate it.
fr, err := newFileReader(lw.layerStore.repository.ctx, lw.bufferedFileWriter.driver, lw.path)
if err != nil {
return "", err
}
tr := io.TeeReader(fr, digester)
if _, err = io.Copy(digestVerifier, tr); err != nil {
return "", err
}
canonical = digester.Digest()
verified = digestVerifier.Verified()
}
if !verified {
context.GetLoggerWithField(lw.layerStore.repository.ctx, "canonical", dgst).
Errorf("canonical digest does match provided digest")
return "", distribution.ErrLayerInvalidDigest{
Digest: dgst,
Reason: fmt.Errorf("content does not match digest"),
}
}
return canonical, nil
}
// moveLayer moves the data into its final, hash-qualified destination,
// identified by dgst. The layer should be validated before commencing the
// move.
func (lw *layerWriter) moveLayer(dgst digest.Digest) error {
blobPath, err := lw.layerStore.repository.pm.path(blobDataPathSpec{
digest: dgst,
})
if err != nil {
return err
}
ctx := lw.layerStore.repository.ctx
// Check for existence
if _, err := lw.driver.Stat(ctx, blobPath); err != nil {
switch err := err.(type) {
case storagedriver.PathNotFoundError:
break // ensure that it doesn't exist.
default:
return err
}
} else {
// If the path exists, we can assume that the content has already
// been uploaded, since the blob storage is content-addressable.
// While it may be corrupted, detection of such corruption belongs
// elsewhere.
return nil
}
// If no data was received, we may not actually have a file on disk. Check
// the size here and write a zero-length file to blobPath if this is the
// case. For the most part, this should only ever happen with zero-length
// tars.
if _, err := lw.driver.Stat(ctx, lw.path); err != nil {
switch err := err.(type) {
case storagedriver.PathNotFoundError:
// HACK(stevvooe): This is slightly dangerous: if we verify above,
// get a hash, then the underlying file is deleted, we risk moving
// a zero-length blob into a nonzero-length blob location. To
// prevent this horrid thing, we employ the hack of only allowing
// to this happen for the zero tarsum.
if dgst == digest.DigestSha256EmptyTar {
return lw.driver.PutContent(ctx, blobPath, []byte{})
}
// We let this fail during the move below.
logrus.
WithField("upload.uuid", lw.UUID()).
WithField("digest", dgst).Warnf("attempted to move zero-length content with non-zero digest")
default:
return err // unrelated error
}
}
return lw.driver.Move(ctx, lw.path, blobPath)
}
// linkLayer links a valid, written layer blob into the registry under the
// named repository for the upload controller.
func (lw *layerWriter) linkLayer(canonical digest.Digest, aliases ...digest.Digest) error {
dgsts := append([]digest.Digest{canonical}, aliases...)
// Don't make duplicate links.
seenDigests := make(map[digest.Digest]struct{}, len(dgsts))
for _, dgst := range dgsts {
if _, seen := seenDigests[dgst]; seen {
continue
}
seenDigests[dgst] = struct{}{}
layerLinkPath, err := lw.layerStore.repository.pm.path(layerLinkPathSpec{
name: lw.layerStore.repository.Name(),
digest: dgst,
})
if err != nil {
return err
}
ctx := lw.layerStore.repository.ctx
if err := lw.layerStore.repository.driver.PutContent(ctx, layerLinkPath, []byte(canonical)); err != nil {
return err
}
}
return nil
}
// removeResources should clean up all resources associated with the upload
// instance. An error will be returned if the clean up cannot proceed. If the
// resources are already not present, no error will be returned.
func (lw *layerWriter) removeResources() error {
dataPath, err := lw.layerStore.repository.pm.path(uploadDataPathSpec{
name: lw.layerStore.repository.Name(),
uuid: lw.uuid,
})
if err != nil {
return err
}
// Resolve and delete the containing directory, which should include any
// upload related files.
dirPath := path.Dir(dataPath)
if err := lw.driver.Delete(lw.layerStore.repository.ctx, dirPath); err != nil {
switch err := err.(type) {
case storagedriver.PathNotFoundError:
break // already gone!
default:
// This should be uncommon enough such that returning an error
// should be okay. At this point, the upload should be mostly
// complete, but perhaps the backend became unaccessible.
logrus.Errorf("unable to delete layer upload resources %q: %v", dirPath, err)
return err
}
}
return nil
}

6
docs/storage/layerwriter_nonresumable.go
View file

@ -1,6 +0,0 @@
// +build noresumabledigest
package storage
func (lw *layerWriter) setupResumableDigester() {
}

9
docs/storage/layerwriter_resumable.go
View file

@ -1,9 +0,0 @@
// +build !noresumabledigest
package storage
import "github.com/docker/distribution/digest"
func (lw *layerWriter) setupResumableDigester() {
lw.resumableDigester = digest.NewCanonicalResumableDigester()
}

258
docs/storage/linkedblobstore.go Normal file
View file

@ -0,0 +1,258 @@
package storage
import (
"net/http"
"time"
"code.google.com/p/go-uuid/uuid"
"github.com/docker/distribution"
"github.com/docker/distribution/context"
"github.com/docker/distribution/digest"
"github.com/docker/distribution/registry/storage/driver"
)
// linkedBlobStore provides a full BlobService that namespaces the blobs to a
// given repository. Effectively, it manages the links in a given repository
// that grant access to the global blob store.
type linkedBlobStore struct {
*blobStore
blobServer distribution.BlobServer
statter distribution.BlobStatter
repository distribution.Repository
ctx context.Context // only to be used where context can't come through method args
// linkPath allows one to control the repository blob link set to which
// the blob store dispatches. This is required because manifest and layer
// blobs have not yet been fully merged. At some point, this functionality
// should be removed an the blob links folder should be merged.
linkPath func(pm *pathMapper, name string, dgst digest.Digest) (string, error)
}
var _ distribution.BlobStore = &linkedBlobStore{}
func (lbs *linkedBlobStore) Stat(ctx context.Context, dgst digest.Digest) (distribution.Descriptor, error) {
return lbs.statter.Stat(ctx, dgst)
}
func (lbs *linkedBlobStore) Get(ctx context.Context, dgst digest.Digest) ([]byte, error) {
canonical, err := lbs.Stat(ctx, dgst) // access check
if err != nil {
return nil, err
}
return lbs.blobStore.Get(ctx, canonical.Digest)
}
func (lbs *linkedBlobStore) Open(ctx context.Context, dgst digest.Digest) (distribution.ReadSeekCloser, error) {
canonical, err := lbs.Stat(ctx, dgst) // access check
if err != nil {
return nil, err
}
return lbs.blobStore.Open(ctx, canonical.Digest)
}
func (lbs *linkedBlobStore) ServeBlob(ctx context.Context, w http.ResponseWriter, r *http.Request, dgst digest.Digest) error {
canonical, err := lbs.Stat(ctx, dgst) // access check
if err != nil {
return err
}
if canonical.MediaType != "" {
// Set the repository local content type.
w.Header().Set("Content-Type", canonical.MediaType)
}
return lbs.blobServer.ServeBlob(ctx, w, r, canonical.Digest)
}
func (lbs *linkedBlobStore) Put(ctx context.Context, mediaType string, p []byte) (distribution.Descriptor, error) {
// Place the data in the blob store first.
desc, err := lbs.blobStore.Put(ctx, mediaType, p)
if err != nil {
context.GetLogger(ctx).Errorf("error putting into main store: %v", err)
return distribution.Descriptor{}, err
}
// TODO(stevvooe): Write out mediatype if incoming differs from what is
// returned by Put above. Note that we should allow updates for a given
// repository.
return desc, lbs.linkBlob(ctx, desc)
}
// Writer begins a blob write session, returning a handle.
func (lbs *linkedBlobStore) Create(ctx context.Context) (distribution.BlobWriter, error) {
context.GetLogger(ctx).Debug("(*linkedBlobStore).Writer")
uuid := uuid.New()
startedAt := time.Now().UTC()
path, err := lbs.blobStore.pm.path(uploadDataPathSpec{
name: lbs.repository.Name(),
id: uuid,
})
if err != nil {
return nil, err
}
startedAtPath, err := lbs.blobStore.pm.path(uploadStartedAtPathSpec{
name: lbs.repository.Name(),
id: uuid,
})
if err != nil {
return nil, err
}
// Write a startedat file for this upload
if err := lbs.blobStore.driver.PutContent(ctx, startedAtPath, []byte(startedAt.Format(time.RFC3339))); err != nil {
return nil, err
}
return lbs.newBlobUpload(ctx, uuid, path, startedAt)
}
func (lbs *linkedBlobStore) Resume(ctx context.Context, id string) (distribution.BlobWriter, error) {
context.GetLogger(ctx).Debug("(*linkedBlobStore).Resume")
startedAtPath, err := lbs.blobStore.pm.path(uploadStartedAtPathSpec{
name: lbs.repository.Name(),
id: id,
})
if err != nil {
return nil, err
}
startedAtBytes, err := lbs.blobStore.driver.GetContent(ctx, startedAtPath)
if err != nil {
switch err := err.(type) {
case driver.PathNotFoundError:
return nil, distribution.ErrBlobUploadUnknown
default:
return nil, err
}
}
startedAt, err := time.Parse(time.RFC3339, string(startedAtBytes))
if err != nil {
return nil, err
}
path, err := lbs.pm.path(uploadDataPathSpec{
name: lbs.repository.Name(),
id: id,
})
if err != nil {
return nil, err
}
return lbs.newBlobUpload(ctx, id, path, startedAt)
}
// newLayerUpload allocates a new upload controller with the given state.
func (lbs *linkedBlobStore) newBlobUpload(ctx context.Context, uuid, path string, startedAt time.Time) (distribution.BlobWriter, error) {
fw, err := newFileWriter(ctx, lbs.driver, path)
if err != nil {
return nil, err
}
bw := &blobWriter{
blobStore: lbs,
id: uuid,
startedAt: startedAt,
bufferedFileWriter: *fw,
}
bw.setupResumableDigester()
return bw, nil
}
// linkBlob links a valid, written blob into the registry under the named
// repository for the upload controller.
func (lbs *linkedBlobStore) linkBlob(ctx context.Context, canonical distribution.Descriptor, aliases ...digest.Digest) error {
dgsts := append([]digest.Digest{canonical.Digest}, aliases...)
// TODO(stevvooe): Need to write out mediatype for only canonical hash
// since we don't care about the aliases. They are generally unused except
// for tarsum but those versions don't care about mediatype.
// Don't make duplicate links.
seenDigests := make(map[digest.Digest]struct{}, len(dgsts))
for _, dgst := range dgsts {
if _, seen := seenDigests[dgst]; seen {
continue
}
seenDigests[dgst] = struct{}{}
blobLinkPath, err := lbs.linkPath(lbs.pm, lbs.repository.Name(), dgst)
if err != nil {
return err
}
if err := lbs.blobStore.link(ctx, blobLinkPath, canonical.Digest); err != nil {
return err
}
}
return nil
}
type linkedBlobStatter struct {
*blobStore
repository distribution.Repository
// linkPath allows one to control the repository blob link set to which
// the blob store dispatches. This is required because manifest and layer
// blobs have not yet been fully merged. At some point, this functionality
// should be removed an the blob links folder should be merged.
linkPath func(pm *pathMapper, name string, dgst digest.Digest) (string, error)
}
var _ distribution.BlobStatter = &linkedBlobStatter{}
func (lbs *linkedBlobStatter) Stat(ctx context.Context, dgst digest.Digest) (distribution.Descriptor, error) {
blobLinkPath, err := lbs.linkPath(lbs.pm, lbs.repository.Name(), dgst)
if err != nil {
return distribution.Descriptor{}, err
}
target, err := lbs.blobStore.readlink(ctx, blobLinkPath)
if err != nil {
switch err := err.(type) {
case driver.PathNotFoundError:
return distribution.Descriptor{}, distribution.ErrBlobUnknown
default:
return distribution.Descriptor{}, err
}
// TODO(stevvooe): For backwards compatibility with data in "_layers", we
// need to hit layerLinkPath, as well. Or, somehow migrate to the new path
// layout.
}
if target != dgst {
// Track when we are doing cross-digest domain lookups. ie, tarsum to sha256.
context.GetLogger(ctx).Warnf("looking up blob with canonical target: %v -> %v", dgst, target)
}
// TODO(stevvooe): Look up repository local mediatype and replace that on
// the returned descriptor.
return lbs.blobStore.statter.Stat(ctx, target)
}
// blobLinkPath provides the path to the blob link, also known as layers.
func blobLinkPath(pm *pathMapper, name string, dgst digest.Digest) (string, error) {
return pm.path(layerLinkPathSpec{name: name, digest: dgst})
}
// manifestRevisionLinkPath provides the path to the manifest revision link.
func manifestRevisionLinkPath(pm *pathMapper, name string, dgst digest.Digest) (string, error) {
return pm.path(layerLinkPathSpec{name: name, digest: dgst})
}

56
docs/storage/manifeststore.go
View file

@ -4,7 +4,7 @@ import (
"fmt"
"github.com/docker/distribution"
ctxu "github.com/docker/distribution/context"
"github.com/docker/distribution/context"
"github.com/docker/distribution/digest"
"github.com/docker/distribution/manifest"
"github.com/docker/libtrust"
@ -12,80 +12,84 @@ import (
type manifestStore struct {
repository *repository
revisionStore *revisionStore
tagStore *tagStore
ctx context.Context
}
var _ distribution.ManifestService = &manifestStore{}
func (ms *manifestStore) Exists(dgst digest.Digest) (bool, error) {
ctxu.GetLogger(ms.repository.ctx).Debug("(*manifestStore).Exists")
return ms.revisionStore.exists(dgst)
context.GetLogger(ms.ctx).Debug("(*manifestStore).Exists")
_, err := ms.revisionStore.blobStore.Stat(ms.ctx, dgst)
if err != nil {
if err == distribution.ErrBlobUnknown {
return false, nil
}
return false, err
}
return true, nil
}
func (ms *manifestStore) Get(dgst digest.Digest) (*manifest.SignedManifest, error) {
ctxu.GetLogger(ms.repository.ctx).Debug("(*manifestStore).Get")
return ms.revisionStore.get(dgst)
context.GetLogger(ms.ctx).Debug("(*manifestStore).Get")
return ms.revisionStore.get(ms.ctx, dgst)
}
func (ms *manifestStore) Put(manifest *manifest.SignedManifest) error {
ctxu.GetLogger(ms.repository.ctx).Debug("(*manifestStore).Put")
// TODO(stevvooe): Add check here to see if the revision is already
// present in the repository. If it is, we should merge the signatures, do
// a shallow verify (or a full one, doesn't matter) and return an error
// indicating what happened.
context.GetLogger(ms.ctx).Debug("(*manifestStore).Put")
// Verify the manifest.
if err := ms.verifyManifest(manifest); err != nil {
if err := ms.verifyManifest(ms.ctx, manifest); err != nil {
return err
}
// Store the revision of the manifest
revision, err := ms.revisionStore.put(manifest)
revision, err := ms.revisionStore.put(ms.ctx, manifest)
if err != nil {
return err
}
// Now, tag the manifest
return ms.tagStore.tag(manifest.Tag, revision)
return ms.tagStore.tag(manifest.Tag, revision.Digest)
}
// Delete removes the revision of the specified manfiest.
func (ms *manifestStore) Delete(dgst digest.Digest) error {
ctxu.GetLogger(ms.repository.ctx).Debug("(*manifestStore).Delete - unsupported")
context.GetLogger(ms.ctx).Debug("(*manifestStore).Delete - unsupported")
return fmt.Errorf("deletion of manifests not supported")
}
func (ms *manifestStore) Tags() ([]string, error) {
ctxu.GetLogger(ms.repository.ctx).Debug("(*manifestStore).Tags")
context.GetLogger(ms.ctx).Debug("(*manifestStore).Tags")
return ms.tagStore.tags()
}
func (ms *manifestStore) ExistsByTag(tag string) (bool, error) {
ctxu.GetLogger(ms.repository.ctx).Debug("(*manifestStore).ExistsByTag")
context.GetLogger(ms.ctx).Debug("(*manifestStore).ExistsByTag")
return ms.tagStore.exists(tag)
}
func (ms *manifestStore) GetByTag(tag string) (*manifest.SignedManifest, error) {
ctxu.GetLogger(ms.repository.ctx).Debug("(*manifestStore).GetByTag")
context.GetLogger(ms.ctx).Debug("(*manifestStore).GetByTag")
dgst, err := ms.tagStore.resolve(tag)
if err != nil {
return nil, err
}
return ms.revisionStore.get(dgst)
return ms.revisionStore.get(ms.ctx, dgst)
}
// verifyManifest ensures that the manifest content is valid from the
// perspective of the registry. It ensures that the signature is valid for the
// enclosed payload. As a policy, the registry only tries to store valid
// content, leaving trust policies of that content up to consumers.
func (ms *manifestStore) verifyManifest(mnfst *manifest.SignedManifest) error {
func (ms *manifestStore) verifyManifest(ctx context.Context, mnfst *manifest.SignedManifest) error {
var errs distribution.ErrManifestVerification
if mnfst.Name != ms.repository.Name() {
// TODO(stevvooe): This needs to be an exported error
errs = append(errs, fmt.Errorf("repository name does not match manifest name"))
}
@ -103,18 +107,18 @@ func (ms *manifestStore) verifyManifest(mnfst *manifest.SignedManifest) error {
}
for _, fsLayer := range mnfst.FSLayers {
exists, err := ms.repository.Layers().Exists(fsLayer.BlobSum)
_, err := ms.repository.Blobs(ctx).Stat(ctx, fsLayer.BlobSum)
if err != nil {
if err != distribution.ErrBlobUnknown {
errs = append(errs, err)
}
if !exists {
errs = append(errs, distribution.ErrUnknownLayer{FSLayer: fsLayer})
// On error here, we always append unknown blob errors.
errs = append(errs, distribution.ErrManifestBlobUnknown{Digest: fsLayer.BlobSum})
}
}
if len(errs) != 0 {
// TODO(stevvooe): These need to be recoverable by a caller.
return errs
}

28
docs/storage/manifeststore_test.go
View file

@ -6,16 +6,15 @@ import (
"reflect"
"testing"
"github.com/docker/distribution/registry/storage/cache"
"github.com/docker/distribution"
"github.com/docker/distribution/context"
"github.com/docker/distribution/digest"
"github.com/docker/distribution/manifest"
"github.com/docker/distribution/registry/storage/cache"
"github.com/docker/distribution/registry/storage/driver"
"github.com/docker/distribution/registry/storage/driver/inmemory"
"github.com/docker/distribution/testutil"
"github.com/docker/libtrust"
"golang.org/x/net/context"
)
type manifestStoreTestEnv struct {
@ -30,7 +29,7 @@ type manifestStoreTestEnv struct {
func newManifestStoreTestEnv(t *testing.T, name, tag string) *manifestStoreTestEnv {
ctx := context.Background()
driver := inmemory.New()
registry := NewRegistryWithDriver(ctx, driver, cache.NewInMemoryLayerInfoCache())
registry := NewRegistryWithDriver(ctx, driver, cache.NewInMemoryBlobDescriptorCacheProvider())
repo, err := registry.Repository(ctx, name)
if err != nil {
@ -108,20 +107,33 @@ func TestManifestStorage(t *testing.T) {
t.Fatalf("expected errors putting manifest")
}
// TODO(stevvooe): We expect errors describing all of the missing layers.
switch err := err.(type) {
case distribution.ErrManifestVerification:
if len(err) != 2 {
t.Fatalf("expected 2 verification errors: %#v", err)
}
for _, err := range err {
if _, ok := err.(distribution.ErrManifestBlobUnknown); !ok {
t.Fatalf("unexpected error type: %v", err)
}
}
default:
t.Fatalf("unexpected error verifying manifest: %v", err)
}
// Now, upload the layers that were missing!
for dgst, rs := range testLayers {
upload, err := env.repository.Layers().Upload()
wr, err := env.repository.Blobs(env.ctx).Create(env.ctx)
if err != nil {
t.Fatalf("unexpected error creating test upload: %v", err)
}
if _, err := io.Copy(upload, rs); err != nil {
if _, err := io.Copy(wr, rs); err != nil {
t.Fatalf("unexpected error copying to upload: %v", err)
}
if _, err := upload.Finish(dgst); err != nil {
if _, err := wr.Commit(env.ctx, distribution.Descriptor{Digest: dgst}); err != nil {
t.Fatalf("unexpected error finishing upload: %v", err)
}
}

44
docs/storage/paths.go
View file

@ -30,7 +30,7 @@ const storagePathVersion = "v2"
// -> <algorithm>/<hex digest>/link
// -> _layers/
// <layer links to blob store>
// -> _uploads/<uuid>
// -> _uploads/<id>
// data
// startedat
// hashstates/<algorithm>/<offset>
@ -47,7 +47,7 @@ const storagePathVersion = "v2"
// is just a directory of layers which are "linked" into a repository. A layer
// can only be accessed through a qualified repository name if it is linked in
// the repository. Uploads of layers are managed in the uploads directory,
// which is key by upload uuid. When all data for an upload is received, the
// which is key by upload id. When all data for an upload is received, the
// data is moved into the blob store and the upload directory is deleted.
// Abandoned uploads can be garbage collected by reading the startedat file
// and removing uploads that have been active for longer than a certain time.
@ -80,20 +80,21 @@ const storagePathVersion = "v2"
// manifestTagIndexEntryPathSpec: <root>/v2/repositories/<name>/_manifests/tags/<tag>/index/<algorithm>/<hex digest>/
// manifestTagIndexEntryLinkPathSpec: <root>/v2/repositories/<name>/_manifests/tags/<tag>/index/<algorithm>/<hex digest>/link
//
// Layers:
// Blobs:
//
// layerLinkPathSpec: <root>/v2/repositories/<name>/_layers/tarsum/<tarsum version>/<tarsum hash alg>/<tarsum hash>/link
// layerLinkPathSpec: <root>/v2/repositories/<name>/_layers/<algorithm>/<hex digest>/link
//
// Uploads:
//
// uploadDataPathSpec: <root>/v2/repositories/<name>/_uploads/<uuid>/data
// uploadStartedAtPathSpec: <root>/v2/repositories/<name>/_uploads/<uuid>/startedat
// uploadHashStatePathSpec: <root>/v2/repositories/<name>/_uploads/<uuid>/hashstates/<algorithm>/<offset>
// uploadDataPathSpec: <root>/v2/repositories/<name>/_uploads/<id>/data
// uploadStartedAtPathSpec: <root>/v2/repositories/<name>/_uploads/<id>/startedat
// uploadHashStatePathSpec: <root>/v2/repositories/<name>/_uploads/<id>/hashstates/<algorithm>/<offset>
//
// Blob Store:
//
// blobPathSpec: <root>/v2/blobs/<algorithm>/<first two hex bytes of digest>/<hex digest>
// blobDataPathSpec: <root>/v2/blobs/<algorithm>/<first two hex bytes of digest>/<hex digest>/data
// blobMediaTypePathSpec: <root>/v2/blobs/<algorithm>/<first two hex bytes of digest>/<hex digest>/data
//
// For more information on the semantic meaning of each path and their
// contents, please see the path spec documentation.
@ -234,9 +235,14 @@ func (pm *pathMapper) path(spec pathSpec) (string, error) {
return "", err
}
layerLinkPathComponents := append(repoPrefix, v.name, "_layers")
// TODO(stevvooe): Right now, all blobs are linked under "_layers". If
// we have future migrations, we may want to rename this to "_blobs".
// A migration strategy would simply leave existing items in place and
// write the new paths, commit a file then delete the old files.
return path.Join(path.Join(append(layerLinkPathComponents, components...)...), "link"), nil
blobLinkPathComponents := append(repoPrefix, v.name, "_layers")
return path.Join(path.Join(append(blobLinkPathComponents, components...)...), "link"), nil
case blobDataPathSpec:
components, err := digestPathComponents(v.digest, true)
if err != nil {
@ -248,15 +254,15 @@ func (pm *pathMapper) path(spec pathSpec) (string, error) {
return path.Join(append(blobPathPrefix, components...)...), nil
case uploadDataPathSpec:
return path.Join(append(repoPrefix, v.name, "_uploads", v.uuid, "data")...), nil
return path.Join(append(repoPrefix, v.name, "_uploads", v.id, "data")...), nil
case uploadStartedAtPathSpec:
return path.Join(append(repoPrefix, v.name, "_uploads", v.uuid, "startedat")...), nil
return path.Join(append(repoPrefix, v.name, "_uploads", v.id, "startedat")...), nil
case uploadHashStatePathSpec:
offset := fmt.Sprintf("%d", v.offset)
if v.list {
offset = "" // Limit to the prefix for listing offsets.
}
return path.Join(append(repoPrefix, v.name, "_uploads", v.uuid, "hashstates", v.alg, offset)...), nil
return path.Join(append(repoPrefix, v.name, "_uploads", v.id, "hashstates", v.alg, offset)...), nil
case repositoriesRootPathSpec:
return path.Join(repoPrefix...), nil
default:
@ -367,8 +373,8 @@ type manifestTagIndexEntryLinkPathSpec struct {
func (manifestTagIndexEntryLinkPathSpec) pathSpec() {}
// layerLink specifies a path for a layer link, which is a file with a blob
// id. The layer link will contain a content addressable blob id reference
// blobLinkPathSpec specifies a path for a blob link, which is a file with a
// blob id. The blob link will contain a content addressable blob id reference
// into the blob store. The format of the contents is as follows:
//
// <algorithm>:<hex digest of layer data>
@ -377,7 +383,7 @@ func (manifestTagIndexEntryLinkPathSpec) pathSpec() {}
//
// sha256:96443a84ce518ac22acb2e985eda402b58ac19ce6f91980bde63726a79d80b36
//
// This says indicates that there is a blob with the id/digest, calculated via
// This indicates that there is a blob with the id/digest, calculated via
// sha256 that can be fetched from the blob store.
type layerLinkPathSpec struct {
name string
@ -415,7 +421,7 @@ func (blobDataPathSpec) pathSpec() {}
// uploads.
type uploadDataPathSpec struct {
name string
uuid string
id string
}
func (uploadDataPathSpec) pathSpec() {}
@ -429,7 +435,7 @@ func (uploadDataPathSpec) pathSpec() {}
// the client to enforce time out policies.
type uploadStartedAtPathSpec struct {
name string
uuid string
id string
}
func (uploadStartedAtPathSpec) pathSpec() {}
@ -437,10 +443,10 @@ func (uploadStartedAtPathSpec) pathSpec() {}
// uploadHashStatePathSpec defines the path parameters for the file that stores
// the hash function state of an upload at a specific byte offset. If `list` is
// set, then the path mapper will generate a list prefix for all hash state
// offsets for the upload identified by the name, uuid, and alg.
// offsets for the upload identified by the name, id, and alg.
type uploadHashStatePathSpec struct {
name string
uuid string
id string
alg string
offset int64
list bool

4
docs/storage/paths_test.go
View file

@ -111,14 +111,14 @@ func TestPathMapper(t *testing.T) {
{
spec: uploadDataPathSpec{
name: "foo/bar",
uuid: "asdf-asdf-asdf-adsf",
id: "asdf-asdf-asdf-adsf",
},
expected: "/pathmapper-test/repositories/foo/bar/_uploads/asdf-asdf-asdf-adsf/data",
},
{
spec: uploadStartedAtPathSpec{
name: "foo/bar",
uuid: "asdf-asdf-asdf-adsf",
id: "asdf-asdf-asdf-adsf",
},
expected: "/pathmapper-test/repositories/foo/bar/_uploads/asdf-asdf-asdf-adsf/startedat",
},

6
docs/storage/purgeuploads_test.go
View file

@ -24,7 +24,7 @@ func testUploadFS(t *testing.T, numUploads int, repoName string, startedAt time.
}
func addUploads(ctx context.Context, t *testing.T, d driver.StorageDriver, uploadID, repo string, startedAt time.Time) {
dataPath, err := pm.path(uploadDataPathSpec{name: repo, uuid: uploadID})
dataPath, err := pm.path(uploadDataPathSpec{name: repo, id: uploadID})
if err != nil {
t.Fatalf("Unable to resolve path")
}
@ -32,7 +32,7 @@ func addUploads(ctx context.Context, t *testing.T, d driver.StorageDriver, uploa
t.Fatalf("Unable to write data file")
}
startedAtPath, err := pm.path(uploadStartedAtPathSpec{name: repo, uuid: uploadID})
startedAtPath, err := pm.path(uploadStartedAtPathSpec{name: repo, id: uploadID})
if err != nil {
t.Fatalf("Unable to resolve path")
}
@ -115,7 +115,7 @@ func TestPurgeOnlyUploads(t *testing.T) {
// Create a directory tree outside _uploads and ensure
// these files aren't deleted.
dataPath, err := pm.path(uploadDataPathSpec{name: "test-repo", uuid: uuid.New()})
dataPath, err := pm.path(uploadDataPathSpec{name: "test-repo", id: uuid.New()})
if err != nil {
t.Fatalf(err.Error())
}

111
docs/storage/registry.go
View file

@ -2,38 +2,53 @@ package storage
import (
"github.com/docker/distribution"
"github.com/docker/distribution/context"
"github.com/docker/distribution/registry/api/v2"
"github.com/docker/distribution/registry/storage/cache"
storagedriver "github.com/docker/distribution/registry/storage/driver"
"golang.org/x/net/context"
)
// registry is the top-level implementation of Registry for use in the storage
// package. All instances should descend from this object.
type registry struct {
driver storagedriver.StorageDriver
pm *pathMapper
blobStore *blobStore
layerInfoCache cache.LayerInfoCache
blobServer distribution.BlobServer
statter distribution.BlobStatter // global statter service.
blobDescriptorCacheProvider cache.BlobDescriptorCacheProvider
}
// NewRegistryWithDriver creates a new registry instance from the provided
// driver. The resulting registry may be shared by multiple goroutines but is
// cheap to allocate.
func NewRegistryWithDriver(ctx context.Context, driver storagedriver.StorageDriver, layerInfoCache cache.LayerInfoCache) distribution.Namespace {
func NewRegistryWithDriver(ctx context.Context, driver storagedriver.StorageDriver, blobDescriptorCacheProvider cache.BlobDescriptorCacheProvider) distribution.Namespace {
// create global statter, with cache.
var statter distribution.BlobStatter = &blobStatter{
driver: driver,
pm: defaultPathMapper,
}
if blobDescriptorCacheProvider != nil {
statter = &cachedBlobStatter{
cache: blobDescriptorCacheProvider,
backend: statter,
}
}
bs := &blobStore{
driver: driver,
pm: defaultPathMapper,
ctx: ctx,
statter: statter,
}
return &registry{
driver: driver,
blobStore: bs,
// TODO(sday): This should be configurable.
pm: defaultPathMapper,
layerInfoCache: layerInfoCache,
blobServer: &blobServer{
driver: driver,
statter: statter,
pathFn: bs.path,
},
blobDescriptorCacheProvider: blobDescriptorCacheProvider,
}
}
@ -54,10 +69,20 @@ func (reg *registry) Repository(ctx context.Context, name string) (distribution.
}
}
var descriptorCache distribution.BlobDescriptorService
if reg.blobDescriptorCacheProvider != nil {
var err error
descriptorCache, err = reg.blobDescriptorCacheProvider.RepositoryScoped(name)
if err != nil {
return nil, err
}
}
return &repository{
ctx: ctx,
registry: reg,
name: name,
descriptorCache: descriptorCache,
}, nil
}
@ -66,6 +91,7 @@ type repository struct {
*registry
ctx context.Context
name string
descriptorCache distribution.BlobDescriptorService
}
// Name returns the name of the repository.
@ -78,47 +104,68 @@ func (repo *repository) Name() string {
// to a request local.
func (repo *repository) Manifests() distribution.ManifestService {
return &manifestStore{
ctx: repo.ctx,
repository: repo,
revisionStore: &revisionStore{
ctx: repo.ctx,
repository: repo,
blobStore: &linkedBlobStore{
ctx: repo.ctx,
blobStore: repo.blobStore,
repository: repo,
statter: &linkedBlobStatter{
blobStore: repo.blobStore,
repository: repo,
linkPath: manifestRevisionLinkPath,
},
// TODO(stevvooe): linkPath limits this blob store to only
// manifests. This instance cannot be used for blob checks.
linkPath: manifestRevisionLinkPath,
},
},
tagStore: &tagStore{
ctx: repo.ctx,
repository: repo,
blobStore: repo.registry.blobStore,
},
}
}
// Layers returns an instance of the LayerService. Instantiation is cheap and
// Blobs returns an instance of the BlobStore. Instantiation is cheap and
// may be context sensitive in the future. The instance should be used similar
// to a request local.
func (repo *repository) Layers() distribution.LayerService {
ls := &layerStore{
repository: repo,
}
if repo.registry.layerInfoCache != nil {
// TODO(stevvooe): This is not the best place to setup a cache. We would
// really like to decouple the cache from the backend but also have the
// manifeset service use the layer service cache. For now, we can simply
// integrate the cache directly. The main issue is that we have layer
// access and layer data coupled in a single object. Work is already under
// way to decouple this.
return &cachedLayerService{
LayerService: ls,
repository: repo,
ctx: repo.ctx,
driver: repo.driver,
func (repo *repository) Blobs(ctx context.Context) distribution.BlobStore {
var statter distribution.BlobStatter = &linkedBlobStatter{
blobStore: repo.blobStore,
cache: repo.registry.layerInfoCache,
repository: repo,
linkPath: blobLinkPath,
}
if repo.descriptorCache != nil {
statter = &cachedBlobStatter{
cache: repo.descriptorCache,
backend: statter,
}
}
return ls
return &linkedBlobStore{
blobStore: repo.blobStore,
blobServer: repo.blobServer,
statter: statter,
repository: repo,
ctx: ctx,
// TODO(stevvooe): linkPath limits this blob store to only layers.
// This instance cannot be used for manifest checks.
linkPath: blobLinkPath,
}
}
func (repo *repository) Signatures() distribution.SignatureService {
return &signatureStore{
repository: repo,
blobStore: repo.blobStore,
ctx: repo.ctx,
}
}

116
docs/storage/revisionstore.go
View file

@ -3,8 +3,8 @@ package storage
import (
"encoding/json"
"github.com/Sirupsen/logrus"
"github.com/docker/distribution"
"github.com/docker/distribution/context"
"github.com/docker/distribution/digest"
"github.com/docker/distribution/manifest"
"github.com/docker/libtrust"
@ -12,47 +12,56 @@ import (
// revisionStore supports storing and managing manifest revisions.
type revisionStore struct {
*repository
repository *repository
blobStore *linkedBlobStore
ctx context.Context
}
// exists returns true if the revision is available in the named repository.
func (rs *revisionStore) exists(revision digest.Digest) (bool, error) {
revpath, err := rs.pm.path(manifestRevisionPathSpec{
name: rs.Name(),
revision: revision,
})
if err != nil {
return false, err
func newRevisionStore(ctx context.Context, repo *repository, blobStore *blobStore) *revisionStore {
return &revisionStore{
ctx: ctx,
repository: repo,
blobStore: &linkedBlobStore{
blobStore: blobStore,
repository: repo,
ctx: ctx,
linkPath: manifestRevisionLinkPath,
},
}
exists, err := exists(rs.repository.ctx, rs.driver, revpath)
if err != nil {
return false, err
}
return exists, nil
}
// get retrieves the manifest, keyed by revision digest.
func (rs *revisionStore) get(revision digest.Digest) (*manifest.SignedManifest, error) {
func (rs *revisionStore) get(ctx context.Context, revision digest.Digest) (*manifest.SignedManifest, error) {
// Ensure that this revision is available in this repository.
if exists, err := rs.exists(revision); err != nil {
return nil, err
} else if !exists {
return nil, distribution.ErrUnknownManifestRevision{
Name: rs.Name(),
_, err := rs.blobStore.Stat(ctx, revision)
if err != nil {
if err == distribution.ErrBlobUnknown {
return nil, distribution.ErrManifestUnknownRevision{
Name: rs.repository.Name(),
Revision: revision,
}
}
content, err := rs.blobStore.get(revision)
return nil, err
}
// TODO(stevvooe): Need to check descriptor from above to ensure that the
// mediatype is as we expect for the manifest store.
content, err := rs.blobStore.Get(ctx, revision)
if err != nil {
if err == distribution.ErrBlobUnknown {
return nil, distribution.ErrManifestUnknownRevision{
Name: rs.repository.Name(),
Revision: revision,
}
}
return nil, err
}
// Fetch the signatures for the manifest
signatures, err := rs.Signatures().Get(revision)
signatures, err := rs.repository.Signatures().Get(revision)
if err != nil {
return nil, err
}
@ -78,69 +87,34 @@ func (rs *revisionStore) get(revision digest.Digest) (*manifest.SignedManifest,
// put stores the manifest in the repository, if not already present. Any
// updated signatures will be stored, as well.
func (rs *revisionStore) put(sm *manifest.SignedManifest) (digest.Digest, error) {
func (rs *revisionStore) put(ctx context.Context, sm *manifest.SignedManifest) (distribution.Descriptor, error) {
// Resolve the payload in the manifest.
payload, err := sm.Payload()
if err != nil {
return "", err
return distribution.Descriptor{}, err
}
// Digest and store the manifest payload in the blob store.
revision, err := rs.blobStore.put(payload)
revision, err := rs.blobStore.Put(ctx, manifest.ManifestMediaType, payload)
if err != nil {
logrus.Errorf("error putting payload into blobstore: %v", err)
return "", err
context.GetLogger(ctx).Errorf("error putting payload into blobstore: %v", err)
return distribution.Descriptor{}, err
}
// Link the revision into the repository.
if err := rs.link(revision); err != nil {
return "", err
if err := rs.blobStore.linkBlob(ctx, revision); err != nil {
return distribution.Descriptor{}, err
}
// Grab each json signature and store them.
signatures, err := sm.Signatures()
if err != nil {
return "", err
return distribution.Descriptor{}, err
}
if err := rs.Signatures().Put(revision, signatures...); err != nil {
return "", err
if err := rs.repository.Signatures().Put(revision.Digest, signatures...); err != nil {
return distribution.Descriptor{}, err
}
return revision, nil
}
// link links the revision into the repository.
func (rs *revisionStore) link(revision digest.Digest) error {
revisionPath, err := rs.pm.path(manifestRevisionLinkPathSpec{
name: rs.Name(),
revision: revision,
})
if err != nil {
return err
}
if exists, err := exists(rs.repository.ctx, rs.driver, revisionPath); err != nil {
return err
} else if exists {
// Revision has already been linked!
return nil
}
return rs.blobStore.link(revisionPath, revision)
}
// delete removes the specified manifest revision from storage.
func (rs *revisionStore) delete(revision digest.Digest) error {
revisionPath, err := rs.pm.path(manifestRevisionPathSpec{
name: rs.Name(),
revision: revision,
})
if err != nil {
return err
}
return rs.driver.Delete(rs.repository.ctx, revisionPath)
}

80
docs/storage/signaturestore.go
View file

@ -10,14 +10,24 @@ import (
)
type signatureStore struct {
*repository
repository *repository
blobStore *blobStore
ctx context.Context
}
func newSignatureStore(ctx context.Context, repo *repository, blobStore *blobStore) *signatureStore {
return &signatureStore{
ctx: ctx,
repository: repo,
blobStore: blobStore,
}
}
var _ distribution.SignatureService = &signatureStore{}
func (s *signatureStore) Get(dgst digest.Digest) ([][]byte, error) {
signaturesPath, err := s.pm.path(manifestSignaturesPathSpec{
name: s.Name(),
signaturesPath, err := s.blobStore.pm.path(manifestSignaturesPathSpec{
name: s.repository.Name(),
revision: dgst,
})
@ -30,7 +40,7 @@ func (s *signatureStore) Get(dgst digest.Digest) ([][]byte, error) {
// can be eliminated by implementing listAll on drivers.
signaturesPath = path.Join(signaturesPath, "sha256")
signaturePaths, err := s.driver.List(s.repository.ctx, signaturesPath)
signaturePaths, err := s.blobStore.driver.List(s.ctx, signaturesPath)
if err != nil {
return nil, err
}
@ -43,27 +53,32 @@ func (s *signatureStore) Get(dgst digest.Digest) ([][]byte, error) {
}
ch := make(chan result)
bs := s.linkedBlobStore(s.ctx, dgst)
for i, sigPath := range signaturePaths {
// Append the link portion
sigPath = path.Join(sigPath, "link")
sigdgst, err := digest.ParseDigest("sha256:" + path.Base(sigPath))
if err != nil {
context.GetLogger(s.ctx).Errorf("could not get digest from path: %q, skipping", sigPath)
continue
}
wg.Add(1)
go func(idx int, sigPath string) {
go func(idx int, sigdgst digest.Digest) {
defer wg.Done()
context.GetLogger(s.ctx).
Debugf("fetching signature from %q", sigPath)
Debugf("fetching signature %q", sigdgst)
r := result{index: idx}
if p, err := s.blobStore.linked(sigPath); err != nil {
if p, err := bs.Get(s.ctx, sigdgst); err != nil {
context.GetLogger(s.ctx).
Errorf("error fetching signature from %q: %v", sigPath, err)
Errorf("error fetching signature %q: %v", sigdgst, err)
r.err = err
} else {
r.signature = p
}
ch <- r
}(i, sigPath)
}(i, sigdgst)
}
done := make(chan struct{})
go func() {
@ -91,25 +106,36 @@ loop:
}
func (s *signatureStore) Put(dgst digest.Digest, signatures ...[]byte) error {
bs := s.linkedBlobStore(s.ctx, dgst)
for _, signature := range signatures {
signatureDigest, err := s.blobStore.put(signature)
if err != nil {
return err
}
signaturePath, err := s.pm.path(manifestSignatureLinkPathSpec{
name: s.Name(),
revision: dgst,
signature: signatureDigest,
})
if err != nil {
return err
}
if err := s.blobStore.link(signaturePath, signatureDigest); err != nil {
if _, err := bs.Put(s.ctx, "application/json", signature); err != nil {
return err
}
}
return nil
}
// namedBlobStore returns the namedBlobStore of the signatures for the
// manifest with the given digest. Effectively, each singature link path
// layout is a unique linked blob store.
func (s *signatureStore) linkedBlobStore(ctx context.Context, revision digest.Digest) *linkedBlobStore {
linkpath := func(pm *pathMapper, name string, dgst digest.Digest) (string, error) {
return pm.path(manifestSignatureLinkPathSpec{
name: name,
revision: revision,
signature: dgst,
})
}
return &linkedBlobStore{
ctx: ctx,
repository: s.repository,
blobStore: s.blobStore,
statter: &linkedBlobStatter{
blobStore: s.blobStore,
repository: s.repository,
linkPath: linkpath,
},
linkPath: linkpath,
}
}

108
docs/storage/tagstore.go
View file

@ -4,31 +4,33 @@ import (
"path"
"github.com/docker/distribution"
// "github.com/docker/distribution/context"
"github.com/docker/distribution/context"
"github.com/docker/distribution/digest"
storagedriver "github.com/docker/distribution/registry/storage/driver"
)
// tagStore provides methods to manage manifest tags in a backend storage driver.
type tagStore struct {
*repository
repository *repository
blobStore *blobStore
ctx context.Context
}
// tags lists the manifest tags for the specified repository.
func (ts *tagStore) tags() ([]string, error) {
p, err := ts.pm.path(manifestTagPathSpec{
name: ts.name,
p, err := ts.blobStore.pm.path(manifestTagPathSpec{
name: ts.repository.Name(),
})
if err != nil {
return nil, err
}
var tags []string
entries, err := ts.driver.List(ts.repository.ctx, p)
entries, err := ts.blobStore.driver.List(ts.ctx, p)
if err != nil {
switch err := err.(type) {
case storagedriver.PathNotFoundError:
return nil, distribution.ErrRepositoryUnknown{Name: ts.name}
return nil, distribution.ErrRepositoryUnknown{Name: ts.repository.Name()}
default:
return nil, err
}
@ -45,15 +47,15 @@ func (ts *tagStore) tags() ([]string, error) {
// exists returns true if the specified manifest tag exists in the repository.
func (ts *tagStore) exists(tag string) (bool, error) {
tagPath, err := ts.pm.path(manifestTagCurrentPathSpec{
name: ts.Name(),
tagPath, err := ts.blobStore.pm.path(manifestTagCurrentPathSpec{
name: ts.repository.Name(),
tag: tag,
})
if err != nil {
return false, err
}
exists, err := exists(ts.repository.ctx, ts.driver, tagPath)
exists, err := exists(ts.ctx, ts.blobStore.driver, tagPath)
if err != nil {
return false, err
}
@ -64,18 +66,8 @@ func (ts *tagStore) exists(tag string) (bool, error) {
// tag tags the digest with the given tag, updating the the store to point at
// the current tag. The digest must point to a manifest.
func (ts *tagStore) tag(tag string, revision digest.Digest) error {
indexEntryPath, err := ts.pm.path(manifestTagIndexEntryLinkPathSpec{
name: ts.Name(),
tag: tag,
revision: revision,
})
if err != nil {
return err
}
currentPath, err := ts.pm.path(manifestTagCurrentPathSpec{
name: ts.Name(),
currentPath, err := ts.blobStore.pm.path(manifestTagCurrentPathSpec{
name: ts.repository.Name(),
tag: tag,
})
@ -83,77 +75,69 @@ func (ts *tagStore) tag(tag string, revision digest.Digest) error {
return err
}
nbs := ts.linkedBlobStore(ts.ctx, tag)
// Link into the index
if err := ts.blobStore.link(indexEntryPath, revision); err != nil {
if err := nbs.linkBlob(ts.ctx, distribution.Descriptor{Digest: revision}); err != nil {
return err
}
// Overwrite the current link
return ts.blobStore.link(currentPath, revision)
return ts.blobStore.link(ts.ctx, currentPath, revision)
}
// resolve the current revision for name and tag.
func (ts *tagStore) resolve(tag string) (digest.Digest, error) {
currentPath, err := ts.pm.path(manifestTagCurrentPathSpec{
name: ts.Name(),
currentPath, err := ts.blobStore.pm.path(manifestTagCurrentPathSpec{
name: ts.repository.Name(),
tag: tag,
})
if err != nil {
return "", err
}
if exists, err := exists(ts.repository.ctx, ts.driver, currentPath); err != nil {
return "", err
} else if !exists {
return "", distribution.ErrManifestUnknown{Name: ts.Name(), Tag: tag}
revision, err := ts.blobStore.readlink(ts.ctx, currentPath)
if err != nil {
switch err.(type) {
case storagedriver.PathNotFoundError:
return "", distribution.ErrManifestUnknown{Name: ts.repository.Name(), Tag: tag}
}
revision, err := ts.blobStore.readlink(currentPath)
if err != nil {
return "", err
}
return revision, nil
}
// revisions returns all revisions with the specified name and tag.
func (ts *tagStore) revisions(tag string) ([]digest.Digest, error) {
manifestTagIndexPath, err := ts.pm.path(manifestTagIndexPathSpec{
name: ts.Name(),
tag: tag,
})
if err != nil {
return nil, err
}
// TODO(stevvooe): Need to append digest alg to get listing of revisions.
manifestTagIndexPath = path.Join(manifestTagIndexPath, "sha256")
entries, err := ts.driver.List(ts.repository.ctx, manifestTagIndexPath)
if err != nil {
return nil, err
}
var revisions []digest.Digest
for _, entry := range entries {
revisions = append(revisions, digest.NewDigestFromHex("sha256", path.Base(entry)))
}
return revisions, nil
}
// delete removes the tag from repository, including the history of all
// revisions that have the specified tag.
func (ts *tagStore) delete(tag string) error {
tagPath, err := ts.pm.path(manifestTagPathSpec{
name: ts.Name(),
tagPath, err := ts.blobStore.pm.path(manifestTagPathSpec{
name: ts.repository.Name(),
tag: tag,
})
if err != nil {
return err
}
return ts.driver.Delete(ts.repository.ctx, tagPath)
return ts.blobStore.driver.Delete(ts.ctx, tagPath)
}
// namedBlobStore returns the namedBlobStore for the named tag, allowing one
// to index manifest blobs by tag name. While the tag store doesn't map
// precisely to the linked blob store, using this ensures the links are
// managed via the same code path.
func (ts *tagStore) linkedBlobStore(ctx context.Context, tag string) *linkedBlobStore {
return &linkedBlobStore{
blobStore: ts.blobStore,
repository: ts.repository,
ctx: ctx,
linkPath: func(pm *pathMapper, name string, dgst digest.Digest) (string, error) {
return pm.path(manifestTagIndexEntryLinkPathSpec{
name: name,
tag: tag,
revision: dgst,
})
},
}
}

21
docs/storage/util.go Normal file
View file

@ -0,0 +1,21 @@
package storage
import (
"github.com/docker/distribution/context"
"github.com/docker/distribution/registry/storage/driver"
)
// Exists provides a utility method to test whether or not a path exists in
// the given driver.
func exists(ctx context.Context, drv driver.StorageDriver, path string) (bool, error) {
if _, err := drv.Stat(ctx, path); err != nil {
switch err := err.(type) {
case driver.PathNotFoundError:
return false, nil
default:
return false, err
}
}
return true, nil
}