distribution/registry/handlers/api_test.go

838 lines
24 KiB
Go
Raw Normal View History

package handlers
import (
2014-11-26 20:16:58 +00:00
"bytes"
"encoding/json"
"fmt"
"io"
"io/ioutil"
"net/http"
"net/http/httptest"
"net/http/httputil"
"net/url"
"os"
"path"
Address server errors received during layer upload This changeset addresses intermittent internal server errors encountered during pushes. The root cause has been isolated to layers that result in identical, empty filesystems but may have some path declarations (imaginge "./"), resulting in different tarsums. The main error message reported during these upload problems was a 500 error, which was not correct. Further investigation showed the errors to be rooted in digest verification when finishing uploads. Inspection of the surrounding code also identified a few issues. PutLayerChunk was slightly refactered into PutLayerUploadComplete. Helper methods were avoided to make handler less confusing. This simplification leveraged an earlier change in the spec that moved non-complete chunk uploads to the PATCH method. Simple logging was also added in the unknown error case that should help to avoid mysterious 500 errors in the future. At the same time, the glaring omission of a proper layer upload cancel method was rectified. This has been added in this change so it is not missed in the future. In the future, we may want to refactor the handler code to be more straightforward, hopefully letting us avoid these problems in the future. Added test cases that reproduce these errors and drove these changes include the following: 1. Push a layer with an empty body results in invalid blob upload. 2. Push a layer with a different tarsum (in this case, empty tar) 3. Deleting a layer upload works. 4. Getting status on a deleted layer upload returns 404. Common functionality was grouped into shared functions to remove repitition. The API tests will still require future love. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-30 05:26:35 +00:00
"reflect"
"strings"
"testing"
"github.com/docker/distribution/configuration"
"github.com/docker/distribution/digest"
"github.com/docker/distribution/manifest"
"github.com/docker/distribution/registry/api/v2"
_ "github.com/docker/distribution/registry/storage/driver/inmemory"
"github.com/docker/distribution/testutil"
"github.com/docker/libtrust"
"github.com/gorilla/handlers"
"golang.org/x/net/context"
)
// TestCheckAPI hits the base endpoint (/v2/) ensures we return the specified
// 200 OK response.
func TestCheckAPI(t *testing.T) {
env := newTestEnv(t)
baseURL, err := env.builder.BuildBaseURL()
if err != nil {
t.Fatalf("unexpected error building base url: %v", err)
}
resp, err := http.Get(baseURL)
if err != nil {
t.Fatalf("unexpected error issuing request: %v", err)
}
defer resp.Body.Close()
checkResponse(t, "issuing api base check", resp, http.StatusOK)
checkHeaders(t, resp, http.Header{
"Content-Type": []string{"application/json; charset=utf-8"},
"Content-Length": []string{"2"},
})
p, err := ioutil.ReadAll(resp.Body)
if err != nil {
t.Fatalf("unexpected error reading response body: %v", err)
}
if string(p) != "{}" {
t.Fatalf("unexpected response body: %v", string(p))
}
}
func TestURLPrefix(t *testing.T) {
config := configuration.Configuration{
Storage: configuration.Storage{
"inmemory": configuration.Parameters{},
},
}
config.HTTP.Prefix = "/test/"
env := newTestEnvWithConfig(t, &config)
baseURL, err := env.builder.BuildBaseURL()
if err != nil {
t.Fatalf("unexpected error building base url: %v", err)
}
parsed, _ := url.Parse(baseURL)
if !strings.HasPrefix(parsed.Path, config.HTTP.Prefix) {
t.Fatalf("Prefix %v not included in test url %v", config.HTTP.Prefix, baseURL)
}
resp, err := http.Get(baseURL)
if err != nil {
t.Fatalf("unexpected error issuing request: %v", err)
}
defer resp.Body.Close()
checkResponse(t, "issuing api base check", resp, http.StatusOK)
checkHeaders(t, resp, http.Header{
"Content-Type": []string{"application/json; charset=utf-8"},
"Content-Length": []string{"2"},
})
}
Refactor Blob Service API 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>
2015-05-12 07:10:29 +00:00
// 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.
env := newTestEnv(t)
imageName := "foo/bar"
// "build" our layer file
layerFile, tarSumStr, err := testutil.CreateRandomTarFile()
if err != nil {
t.Fatalf("error creating random layer file: %v", err)
}
layerDigest := digest.Digest(tarSumStr)
// -----------------------------------
// Test fetch for non-existent content
layerURL, err := env.builder.BuildBlobURL(imageName, layerDigest)
2014-11-26 20:16:58 +00:00
if err != nil {
t.Fatalf("error building url: %v", err)
}
2014-11-26 20:16:58 +00:00
resp, err := http.Get(layerURL)
if err != nil {
t.Fatalf("unexpected error fetching non-existent layer: %v", err)
}
2014-11-26 20:16:58 +00:00
checkResponse(t, "fetching non-existent content", resp, http.StatusNotFound)
// ------------------------------------------
// Test head request for non-existent content
2014-11-26 20:16:58 +00:00
resp, err = http.Head(layerURL)
if err != nil {
t.Fatalf("unexpected error checking head on non-existent layer: %v", err)
}
2014-11-26 20:16:58 +00:00
checkResponse(t, "checking head on non-existent layer", resp, http.StatusNotFound)
// ------------------------------------------
// Start an upload, check the status then cancel
uploadURLBase, uploadUUID := startPushLayer(t, env.builder, imageName)
// A status check should work
resp, err = http.Get(uploadURLBase)
if err != nil {
t.Fatalf("unexpected error getting upload status: %v", err)
}
checkResponse(t, "status of deleted upload", resp, http.StatusNoContent)
checkHeaders(t, resp, http.Header{
"Location": []string{"*"},
"Range": []string{"0-0"},
"Docker-Upload-UUID": []string{uploadUUID},
})
Address server errors received during layer upload This changeset addresses intermittent internal server errors encountered during pushes. The root cause has been isolated to layers that result in identical, empty filesystems but may have some path declarations (imaginge "./"), resulting in different tarsums. The main error message reported during these upload problems was a 500 error, which was not correct. Further investigation showed the errors to be rooted in digest verification when finishing uploads. Inspection of the surrounding code also identified a few issues. PutLayerChunk was slightly refactered into PutLayerUploadComplete. Helper methods were avoided to make handler less confusing. This simplification leveraged an earlier change in the spec that moved non-complete chunk uploads to the PATCH method. Simple logging was also added in the unknown error case that should help to avoid mysterious 500 errors in the future. At the same time, the glaring omission of a proper layer upload cancel method was rectified. This has been added in this change so it is not missed in the future. In the future, we may want to refactor the handler code to be more straightforward, hopefully letting us avoid these problems in the future. Added test cases that reproduce these errors and drove these changes include the following: 1. Push a layer with an empty body results in invalid blob upload. 2. Push a layer with a different tarsum (in this case, empty tar) 3. Deleting a layer upload works. 4. Getting status on a deleted layer upload returns 404. Common functionality was grouped into shared functions to remove repitition. The API tests will still require future love. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-30 05:26:35 +00:00
req, err := http.NewRequest("DELETE", uploadURLBase, nil)
if err != nil {
Address server errors received during layer upload This changeset addresses intermittent internal server errors encountered during pushes. The root cause has been isolated to layers that result in identical, empty filesystems but may have some path declarations (imaginge "./"), resulting in different tarsums. The main error message reported during these upload problems was a 500 error, which was not correct. Further investigation showed the errors to be rooted in digest verification when finishing uploads. Inspection of the surrounding code also identified a few issues. PutLayerChunk was slightly refactered into PutLayerUploadComplete. Helper methods were avoided to make handler less confusing. This simplification leveraged an earlier change in the spec that moved non-complete chunk uploads to the PATCH method. Simple logging was also added in the unknown error case that should help to avoid mysterious 500 errors in the future. At the same time, the glaring omission of a proper layer upload cancel method was rectified. This has been added in this change so it is not missed in the future. In the future, we may want to refactor the handler code to be more straightforward, hopefully letting us avoid these problems in the future. Added test cases that reproduce these errors and drove these changes include the following: 1. Push a layer with an empty body results in invalid blob upload. 2. Push a layer with a different tarsum (in this case, empty tar) 3. Deleting a layer upload works. 4. Getting status on a deleted layer upload returns 404. Common functionality was grouped into shared functions to remove repitition. The API tests will still require future love. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-30 05:26:35 +00:00
t.Fatalf("unexpected error creating delete request: %v", err)
}
Address server errors received during layer upload This changeset addresses intermittent internal server errors encountered during pushes. The root cause has been isolated to layers that result in identical, empty filesystems but may have some path declarations (imaginge "./"), resulting in different tarsums. The main error message reported during these upload problems was a 500 error, which was not correct. Further investigation showed the errors to be rooted in digest verification when finishing uploads. Inspection of the surrounding code also identified a few issues. PutLayerChunk was slightly refactered into PutLayerUploadComplete. Helper methods were avoided to make handler less confusing. This simplification leveraged an earlier change in the spec that moved non-complete chunk uploads to the PATCH method. Simple logging was also added in the unknown error case that should help to avoid mysterious 500 errors in the future. At the same time, the glaring omission of a proper layer upload cancel method was rectified. This has been added in this change so it is not missed in the future. In the future, we may want to refactor the handler code to be more straightforward, hopefully letting us avoid these problems in the future. Added test cases that reproduce these errors and drove these changes include the following: 1. Push a layer with an empty body results in invalid blob upload. 2. Push a layer with a different tarsum (in this case, empty tar) 3. Deleting a layer upload works. 4. Getting status on a deleted layer upload returns 404. Common functionality was grouped into shared functions to remove repitition. The API tests will still require future love. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-30 05:26:35 +00:00
resp, err = http.DefaultClient.Do(req)
if err != nil {
Address server errors received during layer upload This changeset addresses intermittent internal server errors encountered during pushes. The root cause has been isolated to layers that result in identical, empty filesystems but may have some path declarations (imaginge "./"), resulting in different tarsums. The main error message reported during these upload problems was a 500 error, which was not correct. Further investigation showed the errors to be rooted in digest verification when finishing uploads. Inspection of the surrounding code also identified a few issues. PutLayerChunk was slightly refactered into PutLayerUploadComplete. Helper methods were avoided to make handler less confusing. This simplification leveraged an earlier change in the spec that moved non-complete chunk uploads to the PATCH method. Simple logging was also added in the unknown error case that should help to avoid mysterious 500 errors in the future. At the same time, the glaring omission of a proper layer upload cancel method was rectified. This has been added in this change so it is not missed in the future. In the future, we may want to refactor the handler code to be more straightforward, hopefully letting us avoid these problems in the future. Added test cases that reproduce these errors and drove these changes include the following: 1. Push a layer with an empty body results in invalid blob upload. 2. Push a layer with a different tarsum (in this case, empty tar) 3. Deleting a layer upload works. 4. Getting status on a deleted layer upload returns 404. Common functionality was grouped into shared functions to remove repitition. The API tests will still require future love. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-30 05:26:35 +00:00
t.Fatalf("unexpected error sending delete request: %v", err)
}
Address server errors received during layer upload This changeset addresses intermittent internal server errors encountered during pushes. The root cause has been isolated to layers that result in identical, empty filesystems but may have some path declarations (imaginge "./"), resulting in different tarsums. The main error message reported during these upload problems was a 500 error, which was not correct. Further investigation showed the errors to be rooted in digest verification when finishing uploads. Inspection of the surrounding code also identified a few issues. PutLayerChunk was slightly refactered into PutLayerUploadComplete. Helper methods were avoided to make handler less confusing. This simplification leveraged an earlier change in the spec that moved non-complete chunk uploads to the PATCH method. Simple logging was also added in the unknown error case that should help to avoid mysterious 500 errors in the future. At the same time, the glaring omission of a proper layer upload cancel method was rectified. This has been added in this change so it is not missed in the future. In the future, we may want to refactor the handler code to be more straightforward, hopefully letting us avoid these problems in the future. Added test cases that reproduce these errors and drove these changes include the following: 1. Push a layer with an empty body results in invalid blob upload. 2. Push a layer with a different tarsum (in this case, empty tar) 3. Deleting a layer upload works. 4. Getting status on a deleted layer upload returns 404. Common functionality was grouped into shared functions to remove repitition. The API tests will still require future love. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-30 05:26:35 +00:00
checkResponse(t, "deleting upload", resp, http.StatusNoContent)
// A status check should result in 404
resp, err = http.Get(uploadURLBase)
if err != nil {
t.Fatalf("unexpected error getting upload status: %v", err)
}
checkResponse(t, "status of deleted upload", resp, http.StatusNotFound)
// -----------------------------------------
// Do layer push with an empty body and different digest
uploadURLBase, uploadUUID = startPushLayer(t, env.builder, imageName)
resp, err = doPushLayer(t, env.builder, imageName, layerDigest, uploadURLBase, bytes.NewReader([]byte{}))
Address server errors received during layer upload This changeset addresses intermittent internal server errors encountered during pushes. The root cause has been isolated to layers that result in identical, empty filesystems but may have some path declarations (imaginge "./"), resulting in different tarsums. The main error message reported during these upload problems was a 500 error, which was not correct. Further investigation showed the errors to be rooted in digest verification when finishing uploads. Inspection of the surrounding code also identified a few issues. PutLayerChunk was slightly refactered into PutLayerUploadComplete. Helper methods were avoided to make handler less confusing. This simplification leveraged an earlier change in the spec that moved non-complete chunk uploads to the PATCH method. Simple logging was also added in the unknown error case that should help to avoid mysterious 500 errors in the future. At the same time, the glaring omission of a proper layer upload cancel method was rectified. This has been added in this change so it is not missed in the future. In the future, we may want to refactor the handler code to be more straightforward, hopefully letting us avoid these problems in the future. Added test cases that reproduce these errors and drove these changes include the following: 1. Push a layer with an empty body results in invalid blob upload. 2. Push a layer with a different tarsum (in this case, empty tar) 3. Deleting a layer upload works. 4. Getting status on a deleted layer upload returns 404. Common functionality was grouped into shared functions to remove repitition. The API tests will still require future love. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-30 05:26:35 +00:00
if err != nil {
t.Fatalf("unexpected error doing bad layer push: %v", err)
}
Address server errors received during layer upload This changeset addresses intermittent internal server errors encountered during pushes. The root cause has been isolated to layers that result in identical, empty filesystems but may have some path declarations (imaginge "./"), resulting in different tarsums. The main error message reported during these upload problems was a 500 error, which was not correct. Further investigation showed the errors to be rooted in digest verification when finishing uploads. Inspection of the surrounding code also identified a few issues. PutLayerChunk was slightly refactered into PutLayerUploadComplete. Helper methods were avoided to make handler less confusing. This simplification leveraged an earlier change in the spec that moved non-complete chunk uploads to the PATCH method. Simple logging was also added in the unknown error case that should help to avoid mysterious 500 errors in the future. At the same time, the glaring omission of a proper layer upload cancel method was rectified. This has been added in this change so it is not missed in the future. In the future, we may want to refactor the handler code to be more straightforward, hopefully letting us avoid these problems in the future. Added test cases that reproduce these errors and drove these changes include the following: 1. Push a layer with an empty body results in invalid blob upload. 2. Push a layer with a different tarsum (in this case, empty tar) 3. Deleting a layer upload works. 4. Getting status on a deleted layer upload returns 404. Common functionality was grouped into shared functions to remove repitition. The API tests will still require future love. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-30 05:26:35 +00:00
checkResponse(t, "bad layer push", resp, http.StatusBadRequest)
checkBodyHasErrorCodes(t, "bad layer push", resp, v2.ErrorCodeDigestInvalid)
Address server errors received during layer upload This changeset addresses intermittent internal server errors encountered during pushes. The root cause has been isolated to layers that result in identical, empty filesystems but may have some path declarations (imaginge "./"), resulting in different tarsums. The main error message reported during these upload problems was a 500 error, which was not correct. Further investigation showed the errors to be rooted in digest verification when finishing uploads. Inspection of the surrounding code also identified a few issues. PutLayerChunk was slightly refactered into PutLayerUploadComplete. Helper methods were avoided to make handler less confusing. This simplification leveraged an earlier change in the spec that moved non-complete chunk uploads to the PATCH method. Simple logging was also added in the unknown error case that should help to avoid mysterious 500 errors in the future. At the same time, the glaring omission of a proper layer upload cancel method was rectified. This has been added in this change so it is not missed in the future. In the future, we may want to refactor the handler code to be more straightforward, hopefully letting us avoid these problems in the future. Added test cases that reproduce these errors and drove these changes include the following: 1. Push a layer with an empty body results in invalid blob upload. 2. Push a layer with a different tarsum (in this case, empty tar) 3. Deleting a layer upload works. 4. Getting status on a deleted layer upload returns 404. Common functionality was grouped into shared functions to remove repitition. The API tests will still require future love. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-30 05:26:35 +00:00
// -----------------------------------------
// Do layer push with an empty body and correct digest
zeroDigest, err := digest.FromTarArchive(bytes.NewReader([]byte{}))
if err != nil {
t.Fatalf("unexpected error digesting empty buffer: %v", err)
}
Address server errors received during layer upload This changeset addresses intermittent internal server errors encountered during pushes. The root cause has been isolated to layers that result in identical, empty filesystems but may have some path declarations (imaginge "./"), resulting in different tarsums. The main error message reported during these upload problems was a 500 error, which was not correct. Further investigation showed the errors to be rooted in digest verification when finishing uploads. Inspection of the surrounding code also identified a few issues. PutLayerChunk was slightly refactered into PutLayerUploadComplete. Helper methods were avoided to make handler less confusing. This simplification leveraged an earlier change in the spec that moved non-complete chunk uploads to the PATCH method. Simple logging was also added in the unknown error case that should help to avoid mysterious 500 errors in the future. At the same time, the glaring omission of a proper layer upload cancel method was rectified. This has been added in this change so it is not missed in the future. In the future, we may want to refactor the handler code to be more straightforward, hopefully letting us avoid these problems in the future. Added test cases that reproduce these errors and drove these changes include the following: 1. Push a layer with an empty body results in invalid blob upload. 2. Push a layer with a different tarsum (in this case, empty tar) 3. Deleting a layer upload works. 4. Getting status on a deleted layer upload returns 404. Common functionality was grouped into shared functions to remove repitition. The API tests will still require future love. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-30 05:26:35 +00:00
uploadURLBase, uploadUUID = startPushLayer(t, env.builder, imageName)
pushLayer(t, env.builder, imageName, zeroDigest, uploadURLBase, bytes.NewReader([]byte{}))
// -----------------------------------------
// Do layer push with an empty body and correct digest
// This is a valid but empty tarfile!
emptyTar := bytes.Repeat([]byte("\x00"), 1024)
emptyDigest, err := digest.FromTarArchive(bytes.NewReader(emptyTar))
Address server errors received during layer upload This changeset addresses intermittent internal server errors encountered during pushes. The root cause has been isolated to layers that result in identical, empty filesystems but may have some path declarations (imaginge "./"), resulting in different tarsums. The main error message reported during these upload problems was a 500 error, which was not correct. Further investigation showed the errors to be rooted in digest verification when finishing uploads. Inspection of the surrounding code also identified a few issues. PutLayerChunk was slightly refactered into PutLayerUploadComplete. Helper methods were avoided to make handler less confusing. This simplification leveraged an earlier change in the spec that moved non-complete chunk uploads to the PATCH method. Simple logging was also added in the unknown error case that should help to avoid mysterious 500 errors in the future. At the same time, the glaring omission of a proper layer upload cancel method was rectified. This has been added in this change so it is not missed in the future. In the future, we may want to refactor the handler code to be more straightforward, hopefully letting us avoid these problems in the future. Added test cases that reproduce these errors and drove these changes include the following: 1. Push a layer with an empty body results in invalid blob upload. 2. Push a layer with a different tarsum (in this case, empty tar) 3. Deleting a layer upload works. 4. Getting status on a deleted layer upload returns 404. Common functionality was grouped into shared functions to remove repitition. The API tests will still require future love. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-30 05:26:35 +00:00
if err != nil {
t.Fatalf("unexpected error digesting empty tar: %v", err)
Address server errors received during layer upload This changeset addresses intermittent internal server errors encountered during pushes. The root cause has been isolated to layers that result in identical, empty filesystems but may have some path declarations (imaginge "./"), resulting in different tarsums. The main error message reported during these upload problems was a 500 error, which was not correct. Further investigation showed the errors to be rooted in digest verification when finishing uploads. Inspection of the surrounding code also identified a few issues. PutLayerChunk was slightly refactered into PutLayerUploadComplete. Helper methods were avoided to make handler less confusing. This simplification leveraged an earlier change in the spec that moved non-complete chunk uploads to the PATCH method. Simple logging was also added in the unknown error case that should help to avoid mysterious 500 errors in the future. At the same time, the glaring omission of a proper layer upload cancel method was rectified. This has been added in this change so it is not missed in the future. In the future, we may want to refactor the handler code to be more straightforward, hopefully letting us avoid these problems in the future. Added test cases that reproduce these errors and drove these changes include the following: 1. Push a layer with an empty body results in invalid blob upload. 2. Push a layer with a different tarsum (in this case, empty tar) 3. Deleting a layer upload works. 4. Getting status on a deleted layer upload returns 404. Common functionality was grouped into shared functions to remove repitition. The API tests will still require future love. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-30 05:26:35 +00:00
}
uploadURLBase, uploadUUID = startPushLayer(t, env.builder, imageName)
pushLayer(t, env.builder, imageName, emptyDigest, uploadURLBase, bytes.NewReader(emptyTar))
Address server errors received during layer upload This changeset addresses intermittent internal server errors encountered during pushes. The root cause has been isolated to layers that result in identical, empty filesystems but may have some path declarations (imaginge "./"), resulting in different tarsums. The main error message reported during these upload problems was a 500 error, which was not correct. Further investigation showed the errors to be rooted in digest verification when finishing uploads. Inspection of the surrounding code also identified a few issues. PutLayerChunk was slightly refactered into PutLayerUploadComplete. Helper methods were avoided to make handler less confusing. This simplification leveraged an earlier change in the spec that moved non-complete chunk uploads to the PATCH method. Simple logging was also added in the unknown error case that should help to avoid mysterious 500 errors in the future. At the same time, the glaring omission of a proper layer upload cancel method was rectified. This has been added in this change so it is not missed in the future. In the future, we may want to refactor the handler code to be more straightforward, hopefully letting us avoid these problems in the future. Added test cases that reproduce these errors and drove these changes include the following: 1. Push a layer with an empty body results in invalid blob upload. 2. Push a layer with a different tarsum (in this case, empty tar) 3. Deleting a layer upload works. 4. Getting status on a deleted layer upload returns 404. Common functionality was grouped into shared functions to remove repitition. The API tests will still require future love. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-30 05:26:35 +00:00
// ------------------------------------------
// Now, actually do successful upload.
2014-11-26 20:16:58 +00:00
layerLength, _ := layerFile.Seek(0, os.SEEK_END)
layerFile.Seek(0, os.SEEK_SET)
uploadURLBase, uploadUUID = startPushLayer(t, env.builder, imageName)
pushLayer(t, env.builder, imageName, layerDigest, uploadURLBase, layerFile)
2014-11-26 20:16:58 +00:00
// ------------------------------------------
// Now, push just a chunk
layerFile.Seek(0, 0)
Refactor Blob Service API 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>
2015-05-12 07:10:29 +00:00
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)
2014-11-26 20:16:58 +00:00
// ------------------------
// Use a head request to see if the layer exists.
resp, err = http.Head(layerURL)
if err != nil {
t.Fatalf("unexpected error checking head on existing layer: %v", err)
}
2014-11-26 20:16:58 +00:00
checkResponse(t, "checking head on existing layer", resp, http.StatusOK)
checkHeaders(t, resp, http.Header{
"Content-Length": []string{fmt.Sprint(layerLength)},
Refactor Blob Service API 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>
2015-05-12 07:10:29 +00:00
"Docker-Content-Digest": []string{canonicalDigest.String()},
2014-11-26 20:16:58 +00:00
})
// ----------------
// Fetch the layer!
resp, err = http.Get(layerURL)
if err != nil {
t.Fatalf("unexpected error fetching layer: %v", err)
}
2014-11-26 20:16:58 +00:00
checkResponse(t, "fetching layer", resp, http.StatusOK)
checkHeaders(t, resp, http.Header{
"Content-Length": []string{fmt.Sprint(layerLength)},
Refactor Blob Service API 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>
2015-05-12 07:10:29 +00:00
"Docker-Content-Digest": []string{canonicalDigest.String()},
2014-11-26 20:16:58 +00:00
})
2014-11-26 20:16:58 +00:00
// Verify the body
verifier, err := digest.NewDigestVerifier(layerDigest)
if err != nil {
t.Fatalf("unexpected error getting digest verifier: %s", err)
}
2014-11-26 20:16:58 +00:00
io.Copy(verifier, resp.Body)
2014-11-26 20:16:58 +00:00
if !verifier.Verified() {
t.Fatalf("response body did not pass verification")
}
// ----------------
// Fetch the layer with an invalid digest
badURL := strings.Replace(layerURL, "tarsum", "trsum", 1)
resp, err = http.Get(badURL)
if err != nil {
t.Fatalf("unexpected error fetching layer: %v", err)
}
checkResponse(t, "fetching layer bad digest", resp, http.StatusBadRequest)
// Cache headers
resp, err = http.Get(layerURL)
if err != nil {
t.Fatalf("unexpected error fetching layer: %v", err)
}
checkResponse(t, "fetching layer", resp, http.StatusOK)
checkHeaders(t, resp, http.Header{
"Content-Length": []string{fmt.Sprint(layerLength)},
Refactor Blob Service API 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>
2015-05-12 07:10:29 +00:00
"Docker-Content-Digest": []string{canonicalDigest.String()},
"ETag": []string{canonicalDigest.String()},
"Cache-Control": []string{"max-age=31536000"},
})
// Matching etag, gives 304
etag := resp.Header.Get("Etag")
req, err = http.NewRequest("GET", layerURL, nil)
if err != nil {
t.Fatalf("Error constructing request: %s", err)
}
req.Header.Set("If-None-Match", etag)
resp, err = http.DefaultClient.Do(req)
if err != nil {
t.Fatalf("Error constructing request: %s", err)
}
checkResponse(t, "fetching layer with etag", resp, http.StatusNotModified)
// Non-matching etag, gives 200
req, err = http.NewRequest("GET", layerURL, nil)
if err != nil {
t.Fatalf("Error constructing request: %s", err)
}
req.Header.Set("If-None-Match", "")
resp, err = http.DefaultClient.Do(req)
checkResponse(t, "fetching layer with invalid etag", resp, http.StatusOK)
// Missing tests:
// - Upload the same tarsum file under and different repository and
// ensure the content remains uncorrupted.
2014-11-26 20:16:58 +00:00
}
2014-11-26 20:16:58 +00:00
func TestManifestAPI(t *testing.T) {
env := newTestEnv(t)
2014-11-26 20:16:58 +00:00
imageName := "foo/bar"
tag := "thetag"
manifestURL, err := env.builder.BuildManifestURL(imageName, tag)
if err != nil {
2014-11-26 20:16:58 +00:00
t.Fatalf("unexpected error getting manifest url: %v", err)
}
2014-11-26 20:16:58 +00:00
// -----------------------------
// Attempt to fetch the manifest
resp, err := http.Get(manifestURL)
if err != nil {
2014-11-26 20:16:58 +00:00
t.Fatalf("unexpected error getting manifest: %v", err)
}
defer resp.Body.Close()
checkResponse(t, "getting non-existent manifest", resp, http.StatusNotFound)
Address server errors received during layer upload This changeset addresses intermittent internal server errors encountered during pushes. The root cause has been isolated to layers that result in identical, empty filesystems but may have some path declarations (imaginge "./"), resulting in different tarsums. The main error message reported during these upload problems was a 500 error, which was not correct. Further investigation showed the errors to be rooted in digest verification when finishing uploads. Inspection of the surrounding code also identified a few issues. PutLayerChunk was slightly refactered into PutLayerUploadComplete. Helper methods were avoided to make handler less confusing. This simplification leveraged an earlier change in the spec that moved non-complete chunk uploads to the PATCH method. Simple logging was also added in the unknown error case that should help to avoid mysterious 500 errors in the future. At the same time, the glaring omission of a proper layer upload cancel method was rectified. This has been added in this change so it is not missed in the future. In the future, we may want to refactor the handler code to be more straightforward, hopefully letting us avoid these problems in the future. Added test cases that reproduce these errors and drove these changes include the following: 1. Push a layer with an empty body results in invalid blob upload. 2. Push a layer with a different tarsum (in this case, empty tar) 3. Deleting a layer upload works. 4. Getting status on a deleted layer upload returns 404. Common functionality was grouped into shared functions to remove repitition. The API tests will still require future love. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-30 05:26:35 +00:00
checkBodyHasErrorCodes(t, "getting non-existent manifest", resp, v2.ErrorCodeManifestUnknown)
2014-11-26 20:16:58 +00:00
tagsURL, err := env.builder.BuildTagsURL(imageName)
if err != nil {
t.Fatalf("unexpected error building tags url: %v", err)
}
resp, err = http.Get(tagsURL)
if err != nil {
t.Fatalf("unexpected error getting unknown tags: %v", err)
}
defer resp.Body.Close()
// Check that we get an unknown repository error when asking for tags
checkResponse(t, "getting unknown manifest tags", resp, http.StatusNotFound)
Address server errors received during layer upload This changeset addresses intermittent internal server errors encountered during pushes. The root cause has been isolated to layers that result in identical, empty filesystems but may have some path declarations (imaginge "./"), resulting in different tarsums. The main error message reported during these upload problems was a 500 error, which was not correct. Further investigation showed the errors to be rooted in digest verification when finishing uploads. Inspection of the surrounding code also identified a few issues. PutLayerChunk was slightly refactered into PutLayerUploadComplete. Helper methods were avoided to make handler less confusing. This simplification leveraged an earlier change in the spec that moved non-complete chunk uploads to the PATCH method. Simple logging was also added in the unknown error case that should help to avoid mysterious 500 errors in the future. At the same time, the glaring omission of a proper layer upload cancel method was rectified. This has been added in this change so it is not missed in the future. In the future, we may want to refactor the handler code to be more straightforward, hopefully letting us avoid these problems in the future. Added test cases that reproduce these errors and drove these changes include the following: 1. Push a layer with an empty body results in invalid blob upload. 2. Push a layer with a different tarsum (in this case, empty tar) 3. Deleting a layer upload works. 4. Getting status on a deleted layer upload returns 404. Common functionality was grouped into shared functions to remove repitition. The API tests will still require future love. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-30 05:26:35 +00:00
checkBodyHasErrorCodes(t, "getting unknown manifest tags", resp, v2.ErrorCodeNameUnknown)
2014-11-26 20:16:58 +00:00
// --------------------------------
// Attempt to push unsigned manifest with missing layers
unsignedManifest := &manifest.Manifest{
Versioned: manifest.Versioned{
SchemaVersion: 1,
},
2014-11-26 20:16:58 +00:00
Name: imageName,
Tag: tag,
FSLayers: []manifest.FSLayer{
2014-11-26 20:16:58 +00:00
{
BlobSum: "asdf",
},
{
BlobSum: "qwer",
},
},
}
resp = putManifest(t, "putting unsigned manifest", manifestURL, unsignedManifest)
defer resp.Body.Close()
2014-11-26 20:16:58 +00:00
checkResponse(t, "posting unsigned manifest", resp, http.StatusBadRequest)
Address server errors received during layer upload This changeset addresses intermittent internal server errors encountered during pushes. The root cause has been isolated to layers that result in identical, empty filesystems but may have some path declarations (imaginge "./"), resulting in different tarsums. The main error message reported during these upload problems was a 500 error, which was not correct. Further investigation showed the errors to be rooted in digest verification when finishing uploads. Inspection of the surrounding code also identified a few issues. PutLayerChunk was slightly refactered into PutLayerUploadComplete. Helper methods were avoided to make handler less confusing. This simplification leveraged an earlier change in the spec that moved non-complete chunk uploads to the PATCH method. Simple logging was also added in the unknown error case that should help to avoid mysterious 500 errors in the future. At the same time, the glaring omission of a proper layer upload cancel method was rectified. This has been added in this change so it is not missed in the future. In the future, we may want to refactor the handler code to be more straightforward, hopefully letting us avoid these problems in the future. Added test cases that reproduce these errors and drove these changes include the following: 1. Push a layer with an empty body results in invalid blob upload. 2. Push a layer with a different tarsum (in this case, empty tar) 3. Deleting a layer upload works. 4. Getting status on a deleted layer upload returns 404. Common functionality was grouped into shared functions to remove repitition. The API tests will still require future love. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-30 05:26:35 +00:00
_, p, counts := checkBodyHasErrorCodes(t, "getting unknown manifest tags", resp,
v2.ErrorCodeManifestUnverified, v2.ErrorCodeBlobUnknown, v2.ErrorCodeDigestInvalid)
2014-11-26 20:16:58 +00:00
Address server errors received during layer upload This changeset addresses intermittent internal server errors encountered during pushes. The root cause has been isolated to layers that result in identical, empty filesystems but may have some path declarations (imaginge "./"), resulting in different tarsums. The main error message reported during these upload problems was a 500 error, which was not correct. Further investigation showed the errors to be rooted in digest verification when finishing uploads. Inspection of the surrounding code also identified a few issues. PutLayerChunk was slightly refactered into PutLayerUploadComplete. Helper methods were avoided to make handler less confusing. This simplification leveraged an earlier change in the spec that moved non-complete chunk uploads to the PATCH method. Simple logging was also added in the unknown error case that should help to avoid mysterious 500 errors in the future. At the same time, the glaring omission of a proper layer upload cancel method was rectified. This has been added in this change so it is not missed in the future. In the future, we may want to refactor the handler code to be more straightforward, hopefully letting us avoid these problems in the future. Added test cases that reproduce these errors and drove these changes include the following: 1. Push a layer with an empty body results in invalid blob upload. 2. Push a layer with a different tarsum (in this case, empty tar) 3. Deleting a layer upload works. 4. Getting status on a deleted layer upload returns 404. Common functionality was grouped into shared functions to remove repitition. The API tests will still require future love. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-30 05:26:35 +00:00
expectedCounts := map[v2.ErrorCode]int{
v2.ErrorCodeManifestUnverified: 1,
v2.ErrorCodeBlobUnknown: 2,
v2.ErrorCodeDigestInvalid: 2,
2014-11-26 20:16:58 +00:00
}
Address server errors received during layer upload This changeset addresses intermittent internal server errors encountered during pushes. The root cause has been isolated to layers that result in identical, empty filesystems but may have some path declarations (imaginge "./"), resulting in different tarsums. The main error message reported during these upload problems was a 500 error, which was not correct. Further investigation showed the errors to be rooted in digest verification when finishing uploads. Inspection of the surrounding code also identified a few issues. PutLayerChunk was slightly refactered into PutLayerUploadComplete. Helper methods were avoided to make handler less confusing. This simplification leveraged an earlier change in the spec that moved non-complete chunk uploads to the PATCH method. Simple logging was also added in the unknown error case that should help to avoid mysterious 500 errors in the future. At the same time, the glaring omission of a proper layer upload cancel method was rectified. This has been added in this change so it is not missed in the future. In the future, we may want to refactor the handler code to be more straightforward, hopefully letting us avoid these problems in the future. Added test cases that reproduce these errors and drove these changes include the following: 1. Push a layer with an empty body results in invalid blob upload. 2. Push a layer with a different tarsum (in this case, empty tar) 3. Deleting a layer upload works. 4. Getting status on a deleted layer upload returns 404. Common functionality was grouped into shared functions to remove repitition. The API tests will still require future love. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-30 05:26:35 +00:00
if !reflect.DeepEqual(counts, expectedCounts) {
t.Fatalf("unexpected number of error codes encountered: %v\n!=\n%v\n---\n%s", counts, expectedCounts, string(p))
2014-11-26 20:16:58 +00:00
}
2014-12-02 01:40:14 +00:00
// TODO(stevvooe): Add a test case where we take a mostly valid registry,
// tamper with the content and ensure that we get a unverified manifest
// error.
2014-11-26 20:16:58 +00:00
// Push 2 random layers
expectedLayers := make(map[digest.Digest]io.ReadSeeker)
for i := range unsignedManifest.FSLayers {
rs, dgstStr, err := testutil.CreateRandomTarFile()
if err != nil {
2014-11-26 20:16:58 +00:00
t.Fatalf("error creating random layer %d: %v", i, err)
}
2014-11-26 20:16:58 +00:00
dgst := digest.Digest(dgstStr)
2014-11-26 20:16:58 +00:00
expectedLayers[dgst] = rs
unsignedManifest.FSLayers[i].BlobSum = dgst
uploadURLBase, _ := startPushLayer(t, env.builder, imageName)
pushLayer(t, env.builder, imageName, dgst, uploadURLBase, rs)
}
2014-11-26 20:16:58 +00:00
// -------------------
// Push the signed manifest with all layers pushed.
signedManifest, err := manifest.Sign(unsignedManifest, env.pk)
2014-11-26 20:16:58 +00:00
if err != nil {
t.Fatalf("unexpected error signing manifest: %v", err)
}
payload, err := signedManifest.Payload()
checkErr(t, err, "getting manifest payload")
dgst, err := digest.FromBytes(payload)
checkErr(t, err, "digesting manifest")
manifestDigestURL, err := env.builder.BuildManifestURL(imageName, dgst.String())
checkErr(t, err, "building manifest url")
2014-11-26 20:16:58 +00:00
resp = putManifest(t, "putting signed manifest", manifestURL, signedManifest)
checkResponse(t, "putting signed manifest", resp, http.StatusAccepted)
checkHeaders(t, resp, http.Header{
"Location": []string{manifestDigestURL},
"Docker-Content-Digest": []string{dgst.String()},
})
2014-11-26 20:16:58 +00:00
// --------------------
// Push by digest -- should get same result
resp = putManifest(t, "putting signed manifest", manifestDigestURL, signedManifest)
checkResponse(t, "putting signed manifest", resp, http.StatusAccepted)
checkHeaders(t, resp, http.Header{
"Location": []string{manifestDigestURL},
"Docker-Content-Digest": []string{dgst.String()},
})
// ------------------
// Fetch by tag name
2014-11-26 20:16:58 +00:00
resp, err = http.Get(manifestURL)
if err != nil {
t.Fatalf("unexpected error fetching manifest: %v", err)
}
2014-11-26 20:16:58 +00:00
defer resp.Body.Close()
2014-11-26 20:16:58 +00:00
checkResponse(t, "fetching uploaded manifest", resp, http.StatusOK)
checkHeaders(t, resp, http.Header{
"Docker-Content-Digest": []string{dgst.String()},
})
var fetchedManifest manifest.SignedManifest
Address server errors received during layer upload This changeset addresses intermittent internal server errors encountered during pushes. The root cause has been isolated to layers that result in identical, empty filesystems but may have some path declarations (imaginge "./"), resulting in different tarsums. The main error message reported during these upload problems was a 500 error, which was not correct. Further investigation showed the errors to be rooted in digest verification when finishing uploads. Inspection of the surrounding code also identified a few issues. PutLayerChunk was slightly refactered into PutLayerUploadComplete. Helper methods were avoided to make handler less confusing. This simplification leveraged an earlier change in the spec that moved non-complete chunk uploads to the PATCH method. Simple logging was also added in the unknown error case that should help to avoid mysterious 500 errors in the future. At the same time, the glaring omission of a proper layer upload cancel method was rectified. This has been added in this change so it is not missed in the future. In the future, we may want to refactor the handler code to be more straightforward, hopefully letting us avoid these problems in the future. Added test cases that reproduce these errors and drove these changes include the following: 1. Push a layer with an empty body results in invalid blob upload. 2. Push a layer with a different tarsum (in this case, empty tar) 3. Deleting a layer upload works. 4. Getting status on a deleted layer upload returns 404. Common functionality was grouped into shared functions to remove repitition. The API tests will still require future love. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-30 05:26:35 +00:00
dec := json.NewDecoder(resp.Body)
2014-11-26 20:16:58 +00:00
if err := dec.Decode(&fetchedManifest); err != nil {
t.Fatalf("error decoding fetched manifest: %v", err)
}
if !bytes.Equal(fetchedManifest.Raw, signedManifest.Raw) {
t.Fatalf("manifests do not match")
}
// ---------------
// Fetch by digest
resp, err = http.Get(manifestDigestURL)
checkErr(t, err, "fetching manifest by digest")
defer resp.Body.Close()
checkResponse(t, "fetching uploaded manifest", resp, http.StatusOK)
checkHeaders(t, resp, http.Header{
"Docker-Content-Digest": []string{dgst.String()},
})
var fetchedManifestByDigest manifest.SignedManifest
dec = json.NewDecoder(resp.Body)
if err := dec.Decode(&fetchedManifestByDigest); err != nil {
t.Fatalf("error decoding fetched manifest: %v", err)
}
if !bytes.Equal(fetchedManifestByDigest.Raw, signedManifest.Raw) {
t.Fatalf("manifests do not match")
}
// Ensure that the tag is listed.
resp, err = http.Get(tagsURL)
if err != nil {
t.Fatalf("unexpected error getting unknown tags: %v", err)
}
defer resp.Body.Close()
// Check that we get an unknown repository error when asking for tags
checkResponse(t, "getting unknown manifest tags", resp, http.StatusOK)
dec = json.NewDecoder(resp.Body)
var tagsResponse tagsAPIResponse
if err := dec.Decode(&tagsResponse); err != nil {
t.Fatalf("unexpected error decoding error response: %v", err)
}
if tagsResponse.Name != imageName {
t.Fatalf("tags name should match image name: %v != %v", tagsResponse.Name, imageName)
}
if len(tagsResponse.Tags) != 1 {
t.Fatalf("expected some tags in response: %v", tagsResponse.Tags)
}
if tagsResponse.Tags[0] != tag {
t.Fatalf("tag not as expected: %q != %q", tagsResponse.Tags[0], tag)
}
2014-11-26 20:16:58 +00:00
}
type testEnv struct {
pk libtrust.PrivateKey
ctx context.Context
config configuration.Configuration
app *App
server *httptest.Server
builder *v2.URLBuilder
}
func newTestEnv(t *testing.T) *testEnv {
config := configuration.Configuration{
Storage: configuration.Storage{
"inmemory": configuration.Parameters{},
},
}
return newTestEnvWithConfig(t, &config)
}
func newTestEnvWithConfig(t *testing.T, config *configuration.Configuration) *testEnv {
ctx := context.Background()
app := NewApp(ctx, *config)
server := httptest.NewServer(handlers.CombinedLoggingHandler(os.Stderr, app))
builder, err := v2.NewURLBuilderFromString(server.URL + config.HTTP.Prefix)
if err != nil {
t.Fatalf("error creating url builder: %v", err)
}
pk, err := libtrust.GenerateECP256PrivateKey()
if err != nil {
t.Fatalf("unexpected error generating private key: %v", err)
}
return &testEnv{
pk: pk,
ctx: ctx,
config: *config,
app: app,
server: server,
builder: builder,
}
}
2014-11-26 20:16:58 +00:00
func putManifest(t *testing.T, msg, url string, v interface{}) *http.Response {
var body []byte
if sm, ok := v.(*manifest.SignedManifest); ok {
body = sm.Raw
} else {
var err error
body, err = json.MarshalIndent(v, "", " ")
if err != nil {
t.Fatalf("unexpected error marshaling %v: %v", v, err)
}
}
2014-11-26 20:16:58 +00:00
req, err := http.NewRequest("PUT", url, bytes.NewReader(body))
if err != nil {
t.Fatalf("error creating request for %s: %v", msg, err)
}
2014-11-26 20:16:58 +00:00
resp, err := http.DefaultClient.Do(req)
if err != nil {
t.Fatalf("error doing put request while %s: %v", msg, err)
}
2014-11-26 20:16:58 +00:00
return resp
}
func startPushLayer(t *testing.T, ub *v2.URLBuilder, name string) (location string, uuid string) {
layerUploadURL, err := ub.BuildBlobUploadURL(name)
if err != nil {
2014-11-26 20:16:58 +00:00
t.Fatalf("unexpected error building layer upload url: %v", err)
}
2014-11-26 20:16:58 +00:00
resp, err := http.Post(layerUploadURL, "", nil)
if err != nil {
t.Fatalf("unexpected error starting layer push: %v", err)
}
defer resp.Body.Close()
2014-11-26 20:16:58 +00:00
checkResponse(t, fmt.Sprintf("pushing starting layer push %v", name), resp, http.StatusAccepted)
u, err := url.Parse(resp.Header.Get("Location"))
if err != nil {
t.Fatalf("error parsing location header: %v", err)
}
uuid = path.Base(u.Path)
2014-11-26 20:16:58 +00:00
checkHeaders(t, resp, http.Header{
"Location": []string{"*"},
"Content-Length": []string{"0"},
"Docker-Upload-UUID": []string{uuid},
2014-11-26 20:16:58 +00:00
})
return resp.Header.Get("Location"), uuid
2014-11-26 20:16:58 +00:00
}
Address server errors received during layer upload This changeset addresses intermittent internal server errors encountered during pushes. The root cause has been isolated to layers that result in identical, empty filesystems but may have some path declarations (imaginge "./"), resulting in different tarsums. The main error message reported during these upload problems was a 500 error, which was not correct. Further investigation showed the errors to be rooted in digest verification when finishing uploads. Inspection of the surrounding code also identified a few issues. PutLayerChunk was slightly refactered into PutLayerUploadComplete. Helper methods were avoided to make handler less confusing. This simplification leveraged an earlier change in the spec that moved non-complete chunk uploads to the PATCH method. Simple logging was also added in the unknown error case that should help to avoid mysterious 500 errors in the future. At the same time, the glaring omission of a proper layer upload cancel method was rectified. This has been added in this change so it is not missed in the future. In the future, we may want to refactor the handler code to be more straightforward, hopefully letting us avoid these problems in the future. Added test cases that reproduce these errors and drove these changes include the following: 1. Push a layer with an empty body results in invalid blob upload. 2. Push a layer with a different tarsum (in this case, empty tar) 3. Deleting a layer upload works. 4. Getting status on a deleted layer upload returns 404. Common functionality was grouped into shared functions to remove repitition. The API tests will still require future love. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-30 05:26:35 +00:00
// doPushLayer pushes the layer content returning the url on success returning
// the response. If you're only expecting a successful response, use pushLayer.
func doPushLayer(t *testing.T, ub *v2.URLBuilder, name string, dgst digest.Digest, uploadURLBase string, body io.Reader) (*http.Response, error) {
u, err := url.Parse(uploadURLBase)
if err != nil {
t.Fatalf("unexpected error parsing pushLayer url: %v", err)
}
u.RawQuery = url.Values{
"_state": u.Query()["_state"],
2014-11-26 20:16:58 +00:00
"digest": []string{dgst.String()},
}.Encode()
uploadURL := u.String()
2014-11-26 20:16:58 +00:00
// Just do a monolithic upload
Address server errors received during layer upload This changeset addresses intermittent internal server errors encountered during pushes. The root cause has been isolated to layers that result in identical, empty filesystems but may have some path declarations (imaginge "./"), resulting in different tarsums. The main error message reported during these upload problems was a 500 error, which was not correct. Further investigation showed the errors to be rooted in digest verification when finishing uploads. Inspection of the surrounding code also identified a few issues. PutLayerChunk was slightly refactered into PutLayerUploadComplete. Helper methods were avoided to make handler less confusing. This simplification leveraged an earlier change in the spec that moved non-complete chunk uploads to the PATCH method. Simple logging was also added in the unknown error case that should help to avoid mysterious 500 errors in the future. At the same time, the glaring omission of a proper layer upload cancel method was rectified. This has been added in this change so it is not missed in the future. In the future, we may want to refactor the handler code to be more straightforward, hopefully letting us avoid these problems in the future. Added test cases that reproduce these errors and drove these changes include the following: 1. Push a layer with an empty body results in invalid blob upload. 2. Push a layer with a different tarsum (in this case, empty tar) 3. Deleting a layer upload works. 4. Getting status on a deleted layer upload returns 404. Common functionality was grouped into shared functions to remove repitition. The API tests will still require future love. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-30 05:26:35 +00:00
req, err := http.NewRequest("PUT", uploadURL, body)
2014-11-26 20:16:58 +00:00
if err != nil {
t.Fatalf("unexpected error creating new request: %v", err)
}
Address server errors received during layer upload This changeset addresses intermittent internal server errors encountered during pushes. The root cause has been isolated to layers that result in identical, empty filesystems but may have some path declarations (imaginge "./"), resulting in different tarsums. The main error message reported during these upload problems was a 500 error, which was not correct. Further investigation showed the errors to be rooted in digest verification when finishing uploads. Inspection of the surrounding code also identified a few issues. PutLayerChunk was slightly refactered into PutLayerUploadComplete. Helper methods were avoided to make handler less confusing. This simplification leveraged an earlier change in the spec that moved non-complete chunk uploads to the PATCH method. Simple logging was also added in the unknown error case that should help to avoid mysterious 500 errors in the future. At the same time, the glaring omission of a proper layer upload cancel method was rectified. This has been added in this change so it is not missed in the future. In the future, we may want to refactor the handler code to be more straightforward, hopefully letting us avoid these problems in the future. Added test cases that reproduce these errors and drove these changes include the following: 1. Push a layer with an empty body results in invalid blob upload. 2. Push a layer with a different tarsum (in this case, empty tar) 3. Deleting a layer upload works. 4. Getting status on a deleted layer upload returns 404. Common functionality was grouped into shared functions to remove repitition. The API tests will still require future love. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-30 05:26:35 +00:00
return http.DefaultClient.Do(req)
}
// pushLayer pushes the layer content returning the url on success.
func pushLayer(t *testing.T, ub *v2.URLBuilder, name string, dgst digest.Digest, uploadURLBase string, body io.Reader) string {
digester := digest.NewCanonicalDigester()
resp, err := doPushLayer(t, ub, name, dgst, uploadURLBase, io.TeeReader(body, &digester))
2014-11-26 20:16:58 +00:00
if err != nil {
Address server errors received during layer upload This changeset addresses intermittent internal server errors encountered during pushes. The root cause has been isolated to layers that result in identical, empty filesystems but may have some path declarations (imaginge "./"), resulting in different tarsums. The main error message reported during these upload problems was a 500 error, which was not correct. Further investigation showed the errors to be rooted in digest verification when finishing uploads. Inspection of the surrounding code also identified a few issues. PutLayerChunk was slightly refactered into PutLayerUploadComplete. Helper methods were avoided to make handler less confusing. This simplification leveraged an earlier change in the spec that moved non-complete chunk uploads to the PATCH method. Simple logging was also added in the unknown error case that should help to avoid mysterious 500 errors in the future. At the same time, the glaring omission of a proper layer upload cancel method was rectified. This has been added in this change so it is not missed in the future. In the future, we may want to refactor the handler code to be more straightforward, hopefully letting us avoid these problems in the future. Added test cases that reproduce these errors and drove these changes include the following: 1. Push a layer with an empty body results in invalid blob upload. 2. Push a layer with a different tarsum (in this case, empty tar) 3. Deleting a layer upload works. 4. Getting status on a deleted layer upload returns 404. Common functionality was grouped into shared functions to remove repitition. The API tests will still require future love. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-30 05:26:35 +00:00
t.Fatalf("unexpected error doing push layer request: %v", err)
2014-11-26 20:16:58 +00:00
}
defer resp.Body.Close()
2014-11-26 20:16:58 +00:00
checkResponse(t, "putting monolithic chunk", resp, http.StatusCreated)
if err != nil {
t.Fatalf("error generating sha256 digest of body")
}
sha256Dgst := digester.Digest()
expectedLayerURL, err := ub.BuildBlobURL(name, sha256Dgst)
2014-11-26 20:16:58 +00:00
if err != nil {
t.Fatalf("error building expected layer url: %v", err)
}
2014-11-26 20:16:58 +00:00
checkHeaders(t, resp, http.Header{
"Location": []string{expectedLayerURL},
"Content-Length": []string{"0"},
"Docker-Content-Digest": []string{sha256Dgst.String()},
2014-11-26 20:16:58 +00:00
})
return resp.Header.Get("Location")
}
func finishUpload(t *testing.T, ub *v2.URLBuilder, name string, uploadURLBase string, dgst digest.Digest) string {
resp, err := doPushLayer(t, ub, name, dgst, uploadURLBase, nil)
if err != nil {
t.Fatalf("unexpected error doing push layer request: %v", err)
}
defer resp.Body.Close()
checkResponse(t, "putting monolithic chunk", resp, http.StatusCreated)
expectedLayerURL, err := ub.BuildBlobURL(name, dgst)
if err != nil {
t.Fatalf("error building expected layer url: %v", err)
}
checkHeaders(t, resp, http.Header{
"Location": []string{expectedLayerURL},
"Content-Length": []string{"0"},
"Docker-Content-Digest": []string{dgst.String()},
})
return resp.Header.Get("Location")
}
func doPushChunk(t *testing.T, uploadURLBase string, body io.Reader) (*http.Response, digest.Digest, error) {
u, err := url.Parse(uploadURLBase)
if err != nil {
t.Fatalf("unexpected error parsing pushLayer url: %v", err)
}
u.RawQuery = url.Values{
"_state": u.Query()["_state"],
}.Encode()
uploadURL := u.String()
digester := digest.NewCanonicalDigester()
req, err := http.NewRequest("PATCH", uploadURL, io.TeeReader(body, digester))
if err != nil {
t.Fatalf("unexpected error creating new request: %v", err)
}
req.Header.Set("Content-Type", "application/octet-stream")
resp, err := http.DefaultClient.Do(req)
return resp, digester.Digest(), err
}
func pushChunk(t *testing.T, ub *v2.URLBuilder, name string, uploadURLBase string, body io.Reader, length int64) (string, digest.Digest) {
resp, dgst, err := doPushChunk(t, uploadURLBase, body)
if err != nil {
t.Fatalf("unexpected error doing push layer request: %v", err)
}
defer resp.Body.Close()
checkResponse(t, "putting chunk", resp, http.StatusAccepted)
if err != nil {
t.Fatalf("error generating sha256 digest of body")
}
checkHeaders(t, resp, http.Header{
"Range": []string{fmt.Sprintf("0-%d", length-1)},
"Content-Length": []string{"0"},
})
return resp.Header.Get("Location"), dgst
}
2014-11-26 20:16:58 +00:00
func checkResponse(t *testing.T, msg string, resp *http.Response, expectedStatus int) {
if resp.StatusCode != expectedStatus {
t.Logf("unexpected status %s: %v != %v", msg, resp.StatusCode, expectedStatus)
maybeDumpResponse(t, resp)
t.FailNow()
}
}
Address server errors received during layer upload This changeset addresses intermittent internal server errors encountered during pushes. The root cause has been isolated to layers that result in identical, empty filesystems but may have some path declarations (imaginge "./"), resulting in different tarsums. The main error message reported during these upload problems was a 500 error, which was not correct. Further investigation showed the errors to be rooted in digest verification when finishing uploads. Inspection of the surrounding code also identified a few issues. PutLayerChunk was slightly refactered into PutLayerUploadComplete. Helper methods were avoided to make handler less confusing. This simplification leveraged an earlier change in the spec that moved non-complete chunk uploads to the PATCH method. Simple logging was also added in the unknown error case that should help to avoid mysterious 500 errors in the future. At the same time, the glaring omission of a proper layer upload cancel method was rectified. This has been added in this change so it is not missed in the future. In the future, we may want to refactor the handler code to be more straightforward, hopefully letting us avoid these problems in the future. Added test cases that reproduce these errors and drove these changes include the following: 1. Push a layer with an empty body results in invalid blob upload. 2. Push a layer with a different tarsum (in this case, empty tar) 3. Deleting a layer upload works. 4. Getting status on a deleted layer upload returns 404. Common functionality was grouped into shared functions to remove repitition. The API tests will still require future love. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-30 05:26:35 +00:00
// checkBodyHasErrorCodes ensures the body is an error body and has the
// expected error codes, returning the error structure, the json slice and a
// count of the errors by code.
func checkBodyHasErrorCodes(t *testing.T, msg string, resp *http.Response, errorCodes ...v2.ErrorCode) (v2.Errors, []byte, map[v2.ErrorCode]int) {
p, err := ioutil.ReadAll(resp.Body)
if err != nil {
t.Fatalf("unexpected error reading body %s: %v", msg, err)
}
var errs v2.Errors
if err := json.Unmarshal(p, &errs); err != nil {
t.Fatalf("unexpected error decoding error response: %v", err)
}
if len(errs.Errors) == 0 {
t.Fatalf("expected errors in response")
}
// TODO(stevvooe): Shoot. The error setup is not working out. The content-
// type headers are being set after writing the status code.
// if resp.Header.Get("Content-Type") != "application/json; charset=utf-8" {
// t.Fatalf("unexpected content type: %v != 'application/json'",
// resp.Header.Get("Content-Type"))
// }
expected := map[v2.ErrorCode]struct{}{}
counts := map[v2.ErrorCode]int{}
// Initialize map with zeros for expected
for _, code := range errorCodes {
expected[code] = struct{}{}
counts[code] = 0
}
for _, err := range errs.Errors {
if _, ok := expected[err.Code]; !ok {
t.Fatalf("unexpected error code %v encountered during %s: %s ", err.Code, msg, string(p))
Address server errors received during layer upload This changeset addresses intermittent internal server errors encountered during pushes. The root cause has been isolated to layers that result in identical, empty filesystems but may have some path declarations (imaginge "./"), resulting in different tarsums. The main error message reported during these upload problems was a 500 error, which was not correct. Further investigation showed the errors to be rooted in digest verification when finishing uploads. Inspection of the surrounding code also identified a few issues. PutLayerChunk was slightly refactered into PutLayerUploadComplete. Helper methods were avoided to make handler less confusing. This simplification leveraged an earlier change in the spec that moved non-complete chunk uploads to the PATCH method. Simple logging was also added in the unknown error case that should help to avoid mysterious 500 errors in the future. At the same time, the glaring omission of a proper layer upload cancel method was rectified. This has been added in this change so it is not missed in the future. In the future, we may want to refactor the handler code to be more straightforward, hopefully letting us avoid these problems in the future. Added test cases that reproduce these errors and drove these changes include the following: 1. Push a layer with an empty body results in invalid blob upload. 2. Push a layer with a different tarsum (in this case, empty tar) 3. Deleting a layer upload works. 4. Getting status on a deleted layer upload returns 404. Common functionality was grouped into shared functions to remove repitition. The API tests will still require future love. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-30 05:26:35 +00:00
}
counts[err.Code]++
}
// Ensure that counts of expected errors were all non-zero
for code := range expected {
if counts[code] == 0 {
t.Fatalf("expected error code %v not encounterd during %s: %s", code, msg, string(p))
Address server errors received during layer upload This changeset addresses intermittent internal server errors encountered during pushes. The root cause has been isolated to layers that result in identical, empty filesystems but may have some path declarations (imaginge "./"), resulting in different tarsums. The main error message reported during these upload problems was a 500 error, which was not correct. Further investigation showed the errors to be rooted in digest verification when finishing uploads. Inspection of the surrounding code also identified a few issues. PutLayerChunk was slightly refactered into PutLayerUploadComplete. Helper methods were avoided to make handler less confusing. This simplification leveraged an earlier change in the spec that moved non-complete chunk uploads to the PATCH method. Simple logging was also added in the unknown error case that should help to avoid mysterious 500 errors in the future. At the same time, the glaring omission of a proper layer upload cancel method was rectified. This has been added in this change so it is not missed in the future. In the future, we may want to refactor the handler code to be more straightforward, hopefully letting us avoid these problems in the future. Added test cases that reproduce these errors and drove these changes include the following: 1. Push a layer with an empty body results in invalid blob upload. 2. Push a layer with a different tarsum (in this case, empty tar) 3. Deleting a layer upload works. 4. Getting status on a deleted layer upload returns 404. Common functionality was grouped into shared functions to remove repitition. The API tests will still require future love. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-30 05:26:35 +00:00
}
}
return errs, p, counts
}
2014-11-26 20:16:58 +00:00
func maybeDumpResponse(t *testing.T, resp *http.Response) {
if d, err := httputil.DumpResponse(resp, true); err != nil {
t.Logf("error dumping response: %v", err)
} else {
t.Logf("response:\n%s", string(d))
2014-11-26 20:16:58 +00:00
}
}
// matchHeaders checks that the response has at least the headers. If not, the
// test will fail. If a passed in header value is "*", any non-zero value will
// suffice as a match.
func checkHeaders(t *testing.T, resp *http.Response, headers http.Header) {
for k, vs := range headers {
if resp.Header.Get(k) == "" {
t.Fatalf("response missing header %q", k)
}
for _, v := range vs {
if v == "*" {
// Just ensure there is some value.
if len(resp.Header[k]) > 0 {
continue
}
}
for _, hv := range resp.Header[k] {
if hv != v {
t.Fatalf("%v header value not matched in response: %q != %q", k, hv, v)
2014-11-26 20:16:58 +00:00
}
}
}
}
}
func checkErr(t *testing.T, err error, msg string) {
if err != nil {
t.Fatalf("unexpected error %s: %v", msg, err)
}
}