distribution/docs/handlers/app.go

816 lines
27 KiB
Go
Raw Normal View History

package handlers
import (
cryptorand "crypto/rand"
"expvar"
"fmt"
"math/rand"
"net"
"net/http"
"os"
"time"
log "github.com/Sirupsen/logrus"
"github.com/docker/distribution"
"github.com/docker/distribution/configuration"
ctxu "github.com/docker/distribution/context"
"github.com/docker/distribution/health"
"github.com/docker/distribution/notifications"
"github.com/docker/distribution/registry/api/errcode"
"github.com/docker/distribution/registry/api/v2"
"github.com/docker/distribution/registry/auth"
registrymiddleware "github.com/docker/distribution/registry/middleware/registry"
repositorymiddleware "github.com/docker/distribution/registry/middleware/repository"
"github.com/docker/distribution/registry/proxy"
"github.com/docker/distribution/registry/storage"
memorycache "github.com/docker/distribution/registry/storage/cache/memory"
rediscache "github.com/docker/distribution/registry/storage/cache/redis"
storagedriver "github.com/docker/distribution/registry/storage/driver"
"github.com/docker/distribution/registry/storage/driver/factory"
storagemiddleware "github.com/docker/distribution/registry/storage/driver/middleware"
"github.com/garyburd/redigo/redis"
"github.com/gorilla/mux"
"golang.org/x/net/context"
)
// randomSecretSize is the number of random bytes to generate if no secret
// was specified.
const randomSecretSize = 32
// App is a global registry application object. Shared resources can be placed
// on this object that will be accessible from all requests. Any writable
// fields should be protected.
type App struct {
context.Context
Config configuration.Configuration
router *mux.Router // main application router, configured with dispatchers
driver storagedriver.StorageDriver // driver maintains the app global storage driver instance.
registry distribution.Namespace // registry is the primary registry backend for the app instance.
accessController auth.AccessController // main access controller for application
// events contains notification related configuration.
events struct {
sink notifications.Sink
source notifications.SourceRecord
}
redis *redis.Pool
// true if this registry is configured as a pull through cache
isCache bool
}
// NewApp takes a configuration and returns a configured app, ready to serve
// requests. The app only implements ServeHTTP and can be wrapped in other
// handlers accordingly.
func NewApp(ctx context.Context, configuration configuration.Configuration) *App {
app := &App{
Config: configuration,
Context: ctx,
router: v2.RouterWithPrefix(configuration.HTTP.Prefix),
isCache: configuration.Proxy.RemoteURL != "",
}
app.Context = ctxu.WithLogger(app.Context, ctxu.GetLogger(app, "instance.id"))
// Register the handler dispatchers.
app.register(v2.RouteNameBase, func(ctx *Context, r *http.Request) http.Handler {
return http.HandlerFunc(apiBase)
})
app.register(v2.RouteNameManifest, imageManifestDispatcher)
app.register(v2.RouteNameCatalog, catalogDispatcher)
app.register(v2.RouteNameTags, tagsDispatcher)
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
app.register(v2.RouteNameBlob, blobDispatcher)
app.register(v2.RouteNameBlobUpload, blobUploadDispatcher)
app.register(v2.RouteNameBlobUploadChunk, blobUploadDispatcher)
var err error
app.driver, err = factory.Create(configuration.Storage.Type(), configuration.Storage.Parameters())
if err != nil {
// TODO(stevvooe): Move the creation of a service into a protected
// method, where this is created lazily. Its status can be queried via
// a health check.
panic(err)
}
purgeConfig := uploadPurgeDefaultConfig()
if mc, ok := configuration.Storage["maintenance"]; ok {
for k, v := range mc {
switch k {
case "uploadpurging":
purgeConfig = v.(map[interface{}]interface{})
}
}
}
startUploadPurger(app, app.driver, ctxu.GetLogger(app), purgeConfig)
app.driver, err = applyStorageMiddleware(app.driver, configuration.Middleware["storage"])
if err != nil {
panic(err)
}
app.configureSecret(&configuration)
app.configureEvents(&configuration)
app.configureRedis(&configuration)
app.configureLogHook(&configuration)
// configure deletion
var deleteEnabled bool
if d, ok := configuration.Storage["delete"]; ok {
e, ok := d["enabled"]
if ok {
if deleteEnabled, ok = e.(bool); !ok {
deleteEnabled = false
}
}
}
// configure redirects
var redirectDisabled bool
if redirectConfig, ok := configuration.Storage["redirect"]; ok {
v := redirectConfig["disable"]
switch v := v.(type) {
case bool:
redirectDisabled = v
default:
panic(fmt.Sprintf("invalid type for redirect config: %#v", redirectConfig))
}
if redirectDisabled {
ctxu.GetLogger(app).Infof("backend redirection disabled")
}
}
// configure storage caches
if cc, ok := configuration.Storage["cache"]; ok {
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
v, ok := cc["blobdescriptor"]
if !ok {
// Backwards compatible: "layerinfo" == "blobdescriptor"
v = cc["layerinfo"]
}
switch v {
case "redis":
if app.redis == nil {
panic("redis configuration required to use for layerinfo cache")
}
app.registry = storage.NewRegistryWithDriver(app, app.driver, rediscache.NewRedisBlobDescriptorCacheProvider(app.redis), deleteEnabled, !redirectDisabled, app.isCache)
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
ctxu.GetLogger(app).Infof("using redis blob descriptor cache")
case "inmemory":
app.registry = storage.NewRegistryWithDriver(app, app.driver, memorycache.NewInMemoryBlobDescriptorCacheProvider(), deleteEnabled, !redirectDisabled, app.isCache)
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
ctxu.GetLogger(app).Infof("using inmemory blob descriptor cache")
default:
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
if v != "" {
ctxu.GetLogger(app).Warnf("unknown cache type %q, caching disabled", configuration.Storage["cache"])
}
}
}
if app.registry == nil {
// configure the registry if no cache section is available.
app.registry = storage.NewRegistryWithDriver(app.Context, app.driver, nil, deleteEnabled, !redirectDisabled, app.isCache)
}
app.registry, err = applyRegistryMiddleware(app.Context, app.registry, configuration.Middleware["registry"])
if err != nil {
panic(err)
}
authType := configuration.Auth.Type()
if authType != "" {
accessController, err := auth.GetAccessController(configuration.Auth.Type(), configuration.Auth.Parameters())
if err != nil {
panic(fmt.Sprintf("unable to configure authorization (%s): %v", authType, err))
}
app.accessController = accessController
ctxu.GetLogger(app).Debugf("configured %q access controller", authType)
}
// configure as a pull through cache
if configuration.Proxy.RemoteURL != "" {
app.registry, err = proxy.NewRegistryPullThroughCache(ctx, app.registry, app.driver, configuration.Proxy)
if err != nil {
panic(err.Error())
}
app.isCache = true
ctxu.GetLogger(app).Info("Registry configured as a proxy cache to ", configuration.Proxy.RemoteURL)
}
return app
}
// RegisterHealthChecks is an awful hack to defer health check registration
// control to callers. This should only ever be called once per registry
// process, typically in a main function. The correct way would be register
// health checks outside of app, since multiple apps may exist in the same
// process. Because the configuration and app are tightly coupled,
// implementing this properly will require a refactor. This method may panic
// if called twice in the same process.
func (app *App) RegisterHealthChecks() {
health.RegisterPeriodicThresholdFunc("storagedriver_"+app.Config.Storage.Type(), 10*time.Second, 3, func() error {
_, err := app.driver.List(app, "/") // "/" should always exist
return err // any error will be treated as failure
})
}
// register a handler with the application, by route name. The handler will be
// passed through the application filters and context will be constructed at
// request time.
func (app *App) register(routeName string, dispatch dispatchFunc) {
// TODO(stevvooe): This odd dispatcher/route registration is by-product of
// some limitations in the gorilla/mux router. We are using it to keep
// routing consistent between the client and server, but we may want to
// replace it with manual routing and structure-based dispatch for better
// control over the request execution.
app.router.GetRoute(routeName).Handler(app.dispatcher(dispatch))
}
// configureEvents prepares the event sink for action.
func (app *App) configureEvents(configuration *configuration.Configuration) {
// Configure all of the endpoint sinks.
var sinks []notifications.Sink
for _, endpoint := range configuration.Notifications.Endpoints {
if endpoint.Disabled {
ctxu.GetLogger(app).Infof("endpoint %s disabled, skipping", endpoint.Name)
continue
}
ctxu.GetLogger(app).Infof("configuring endpoint %v (%v), timeout=%s, headers=%v", endpoint.Name, endpoint.URL, endpoint.Timeout, endpoint.Headers)
endpoint := notifications.NewEndpoint(endpoint.Name, endpoint.URL, notifications.EndpointConfig{
Timeout: endpoint.Timeout,
Threshold: endpoint.Threshold,
Backoff: endpoint.Backoff,
Headers: endpoint.Headers,
})
sinks = append(sinks, endpoint)
}
// NOTE(stevvooe): Moving to a new queueing implementation is as easy as
// replacing broadcaster with a rabbitmq implementation. It's recommended
// that the registry instances also act as the workers to keep deployment
// simple.
app.events.sink = notifications.NewBroadcaster(sinks...)
// Populate registry event source
hostname, err := os.Hostname()
if err != nil {
hostname = configuration.HTTP.Addr
} else {
// try to pick the port off the config
_, port, err := net.SplitHostPort(configuration.HTTP.Addr)
if err == nil {
hostname = net.JoinHostPort(hostname, port)
}
}
app.events.source = notifications.SourceRecord{
Addr: hostname,
InstanceID: ctxu.GetStringValue(app, "instance.id"),
}
}
func (app *App) configureRedis(configuration *configuration.Configuration) {
if configuration.Redis.Addr == "" {
ctxu.GetLogger(app).Infof("redis not configured")
return
}
pool := &redis.Pool{
Dial: func() (redis.Conn, error) {
// TODO(stevvooe): Yet another use case for contextual timing.
ctx := context.WithValue(app, "redis.connect.startedat", time.Now())
done := func(err error) {
logger := ctxu.GetLoggerWithField(ctx, "redis.connect.duration",
ctxu.Since(ctx, "redis.connect.startedat"))
if err != nil {
logger.Errorf("redis: error connecting: %v", err)
} else {
logger.Infof("redis: connect %v", configuration.Redis.Addr)
}
}
conn, err := redis.DialTimeout("tcp",
configuration.Redis.Addr,
configuration.Redis.DialTimeout,
configuration.Redis.ReadTimeout,
configuration.Redis.WriteTimeout)
if err != nil {
ctxu.GetLogger(app).Errorf("error connecting to redis instance %s: %v",
configuration.Redis.Addr, err)
done(err)
return nil, err
}
// authorize the connection
if configuration.Redis.Password != "" {
if _, err = conn.Do("AUTH", configuration.Redis.Password); err != nil {
defer conn.Close()
done(err)
return nil, err
}
}
// select the database to use
if configuration.Redis.DB != 0 {
if _, err = conn.Do("SELECT", configuration.Redis.DB); err != nil {
defer conn.Close()
done(err)
return nil, err
}
}
done(nil)
return conn, nil
},
MaxIdle: configuration.Redis.Pool.MaxIdle,
MaxActive: configuration.Redis.Pool.MaxActive,
IdleTimeout: configuration.Redis.Pool.IdleTimeout,
TestOnBorrow: func(c redis.Conn, t time.Time) error {
// TODO(stevvooe): We can probably do something more interesting
// here with the health package.
_, err := c.Do("PING")
return err
},
Wait: false, // if a connection is not avialable, proceed without cache.
}
app.redis = pool
// setup expvar
registry := expvar.Get("registry")
if registry == nil {
registry = expvar.NewMap("registry")
}
registry.(*expvar.Map).Set("redis", expvar.Func(func() interface{} {
return map[string]interface{}{
"Config": configuration.Redis,
"Active": app.redis.ActiveCount(),
}
}))
}
// configureLogHook prepares logging hook parameters.
func (app *App) configureLogHook(configuration *configuration.Configuration) {
entry, ok := ctxu.GetLogger(app).(*log.Entry)
if !ok {
// somehow, we are not using logrus
return
}
logger := entry.Logger
for _, configHook := range configuration.Log.Hooks {
if !configHook.Disabled {
switch configHook.Type {
case "mail":
hook := &logHook{}
hook.LevelsParam = configHook.Levels
hook.Mail = &mailer{
Addr: configHook.MailOptions.SMTP.Addr,
Username: configHook.MailOptions.SMTP.Username,
Password: configHook.MailOptions.SMTP.Password,
Insecure: configHook.MailOptions.SMTP.Insecure,
From: configHook.MailOptions.From,
To: configHook.MailOptions.To,
}
logger.Hooks.Add(hook)
default:
}
}
}
}
// configureSecret creates a random secret if a secret wasn't included in the
// configuration.
func (app *App) configureSecret(configuration *configuration.Configuration) {
if configuration.HTTP.Secret == "" {
var secretBytes [randomSecretSize]byte
if _, err := cryptorand.Read(secretBytes[:]); err != nil {
panic(fmt.Sprintf("could not generate random bytes for HTTP secret: %v", err))
}
configuration.HTTP.Secret = string(secretBytes[:])
ctxu.GetLogger(app).Warn("No HTTP secret provided - generated random secret. This may cause problems with uploads if multiple registries are behind a load-balancer. To provide a shared secret, fill in http.secret in the configuration file or set the REGISTRY_HTTP_SECRET environment variable.")
}
}
func (app *App) ServeHTTP(w http.ResponseWriter, r *http.Request) {
defer r.Body.Close() // ensure that request body is always closed.
// Instantiate an http context here so we can track the error codes
// returned by the request router.
ctx := defaultContextManager.context(app, w, r)
defer func() {
status, ok := ctx.Value("http.response.status").(int)
if ok && status >= 200 && status <= 399 {
ctxu.GetResponseLogger(ctx).Infof("response completed")
}
}()
defer defaultContextManager.release(ctx)
// NOTE(stevvooe): Total hack to get instrumented responsewriter from context.
var err error
w, err = ctxu.GetResponseWriter(ctx)
if err != nil {
ctxu.GetLogger(ctx).Warnf("response writer not found in context")
}
// Set a header with the Docker Distribution API Version for all responses.
w.Header().Add("Docker-Distribution-API-Version", "registry/2.0")
app.router.ServeHTTP(w, r)
}
// dispatchFunc takes a context and request and returns a constructed handler
// for the route. The dispatcher will use this to dynamically create request
// specific handlers for each endpoint without creating a new router for each
// request.
type dispatchFunc func(ctx *Context, r *http.Request) http.Handler
// TODO(stevvooe): dispatchers should probably have some validation error
// chain with proper error reporting.
// dispatcher returns a handler that constructs a request specific context and
// handler, using the dispatch factory function.
func (app *App) dispatcher(dispatch dispatchFunc) http.Handler {
return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
context := app.context(w, r)
if err := app.authorized(w, r, context); err != nil {
ctxu.GetLogger(context).Warnf("error authorizing context: %v", err)
return
}
2015-04-14 23:07:23 +00:00
// Add username to request logging
context.Context = ctxu.WithLogger(context.Context, ctxu.GetLogger(context.Context, "auth.user.name"))
if app.nameRequired(r) {
repository, err := app.registry.Repository(context, getName(context))
if err != nil {
ctxu.GetLogger(context).Errorf("error resolving repository: %v", err)
switch err := err.(type) {
case distribution.ErrRepositoryUnknown:
context.Errors = append(context.Errors, v2.ErrorCodeNameUnknown.WithDetail(err))
case distribution.ErrRepositoryNameInvalid:
context.Errors = append(context.Errors, v2.ErrorCodeNameInvalid.WithDetail(err))
}
if err := errcode.ServeJSON(w, context.Errors); err != nil {
ctxu.GetLogger(context).Errorf("error serving error json: %v (from %v)", err, context.Errors)
}
return
}
// assign and decorate the authorized repository with an event bridge.
context.Repository = notifications.Listen(
repository,
app.eventBridge(context, r))
context.Repository, err = applyRepoMiddleware(context.Context, context.Repository, app.Config.Middleware["repository"])
if err != nil {
ctxu.GetLogger(context).Errorf("error initializing repository middleware: %v", err)
context.Errors = append(context.Errors, errcode.ErrorCodeUnknown.WithDetail(err))
if err := errcode.ServeJSON(w, context.Errors); err != nil {
ctxu.GetLogger(context).Errorf("error serving error json: %v (from %v)", err, context.Errors)
}
return
}
}
dispatch(context, r).ServeHTTP(w, r)
// Automated error response handling here. Handlers may return their
// own errors if they need different behavior (such as range errors
// for layer upload).
if context.Errors.Len() > 0 {
if err := errcode.ServeJSON(w, context.Errors); err != nil {
ctxu.GetLogger(context).Errorf("error serving error json: %v (from %v)", err, context.Errors)
}
app.logError(context, context.Errors)
}
})
}
func (app *App) logError(context context.Context, errors errcode.Errors) {
for _, e1 := range errors {
var c ctxu.Context
switch e1.(type) {
case errcode.Error:
e, _ := e1.(errcode.Error)
c = ctxu.WithValue(context, "err.code", e.Code)
c = ctxu.WithValue(c, "err.message", e.Code.Message())
c = ctxu.WithValue(c, "err.detail", e.Detail)
case errcode.ErrorCode:
e, _ := e1.(errcode.ErrorCode)
c = ctxu.WithValue(context, "err.code", e)
c = ctxu.WithValue(c, "err.message", e.Message())
default:
// just normal go 'error'
c = ctxu.WithValue(context, "err.code", errcode.ErrorCodeUnknown)
c = ctxu.WithValue(c, "err.message", e1.Error())
}
c = ctxu.WithLogger(c, ctxu.GetLogger(c,
"err.code",
"err.message",
"err.detail"))
ctxu.GetResponseLogger(c).Errorf("response completed with error")
}
}
// context constructs the context object for the application. This only be
// called once per request.
func (app *App) context(w http.ResponseWriter, r *http.Request) *Context {
ctx := defaultContextManager.context(app, w, r)
ctx = ctxu.WithVars(ctx, r)
ctx = ctxu.WithLogger(ctx, ctxu.GetLogger(ctx,
"vars.name",
"vars.reference",
"vars.digest",
"vars.uuid"))
context := &Context{
App: app,
Context: ctx,
urlBuilder: v2.NewURLBuilderFromRequest(r),
}
return context
}
// authorized checks if the request can proceed with access to the requested
// repository. If it succeeds, the context may access the requested
// repository. An error will be returned if access is not available.
func (app *App) authorized(w http.ResponseWriter, r *http.Request, context *Context) error {
ctxu.GetLogger(context).Debug("authorizing request")
repo := getName(context)
if app.accessController == nil {
return nil // access controller is not enabled.
}
var accessRecords []auth.Access
if repo != "" {
accessRecords = appendAccessRecords(accessRecords, r.Method, repo)
} else {
// Only allow the name not to be set on the base route.
if app.nameRequired(r) {
// For this to be properly secured, repo must always be set for a
// resource that may make a modification. The only condition under
// which name is not set and we still allow access is when the
// base route is accessed. This section prevents us from making
// that mistake elsewhere in the code, allowing any operation to
// proceed.
if err := errcode.ServeJSON(w, errcode.ErrorCodeUnauthorized); err != nil {
ctxu.GetLogger(context).Errorf("error serving error json: %v (from %v)", err, context.Errors)
}
return fmt.Errorf("forbidden: no repository name")
}
accessRecords = appendCatalogAccessRecord(accessRecords, r)
}
ctx, err := app.accessController.Authorized(context.Context, accessRecords...)
if err != nil {
switch err := err.(type) {
case auth.Challenge:
// Add the appropriate WWW-Auth header
err.SetHeaders(w)
if err := errcode.ServeJSON(w, errcode.ErrorCodeUnauthorized.WithDetail(accessRecords)); err != nil {
ctxu.GetLogger(context).Errorf("error serving error json: %v (from %v)", err, context.Errors)
}
default:
// This condition is a potential security problem either in
// the configuration or whatever is backing the access
// controller. Just return a bad request with no information
// to avoid exposure. The request should not proceed.
ctxu.GetLogger(context).Errorf("error checking authorization: %v", err)
w.WriteHeader(http.StatusBadRequest)
}
return err
}
// TODO(stevvooe): This pattern needs to be cleaned up a bit. One context
// should be replaced by another, rather than replacing the context on a
// mutable object.
context.Context = ctx
return nil
}
// eventBridge returns a bridge for the current request, configured with the
// correct actor and source.
func (app *App) eventBridge(ctx *Context, r *http.Request) notifications.Listener {
actor := notifications.ActorRecord{
Name: getUserName(ctx, r),
}
request := notifications.NewRequestRecord(ctxu.GetRequestID(ctx), r)
return notifications.NewBridge(ctx.urlBuilder, app.events.source, actor, request, app.events.sink)
}
// nameRequired returns true if the route requires a name.
func (app *App) nameRequired(r *http.Request) bool {
route := mux.CurrentRoute(r)
routeName := route.GetName()
return route == nil || (routeName != v2.RouteNameBase && routeName != v2.RouteNameCatalog)
}
// apiBase implements a simple yes-man for doing overall checks against the
// api. This can support auth roundtrips to support docker login.
func apiBase(w http.ResponseWriter, r *http.Request) {
const emptyJSON = "{}"
// Provide a simple /v2/ 200 OK response with empty json response.
w.Header().Set("Content-Type", "application/json; charset=utf-8")
w.Header().Set("Content-Length", fmt.Sprint(len(emptyJSON)))
fmt.Fprint(w, emptyJSON)
}
// appendAccessRecords checks the method and adds the appropriate Access records to the records list.
func appendAccessRecords(records []auth.Access, method string, repo string) []auth.Access {
resource := auth.Resource{
Type: "repository",
Name: repo,
}
switch method {
case "GET", "HEAD":
records = append(records,
auth.Access{
Resource: resource,
Action: "pull",
})
case "POST", "PUT", "PATCH":
records = append(records,
auth.Access{
Resource: resource,
Action: "pull",
},
auth.Access{
Resource: resource,
Action: "push",
})
case "DELETE":
// DELETE access requires full admin rights, which is represented
// as "*". This may not be ideal.
records = append(records,
auth.Access{
Resource: resource,
Action: "*",
})
}
return records
}
// Add the access record for the catalog if it's our current route
func appendCatalogAccessRecord(accessRecords []auth.Access, r *http.Request) []auth.Access {
route := mux.CurrentRoute(r)
routeName := route.GetName()
if routeName == v2.RouteNameCatalog {
resource := auth.Resource{
Type: "registry",
Name: "catalog",
}
accessRecords = append(accessRecords,
auth.Access{
Resource: resource,
Action: "*",
})
}
return accessRecords
}
// applyRegistryMiddleware wraps a registry instance with the configured middlewares
func applyRegistryMiddleware(ctx context.Context, registry distribution.Namespace, middlewares []configuration.Middleware) (distribution.Namespace, error) {
for _, mw := range middlewares {
rmw, err := registrymiddleware.Get(ctx, mw.Name, mw.Options, registry)
if err != nil {
return nil, fmt.Errorf("unable to configure registry middleware (%s): %s", mw.Name, err)
}
registry = rmw
}
return registry, nil
}
// applyRepoMiddleware wraps a repository with the configured middlewares
func applyRepoMiddleware(ctx context.Context, repository distribution.Repository, middlewares []configuration.Middleware) (distribution.Repository, error) {
for _, mw := range middlewares {
rmw, err := repositorymiddleware.Get(ctx, mw.Name, mw.Options, repository)
if err != nil {
return nil, err
}
repository = rmw
}
return repository, nil
}
// applyStorageMiddleware wraps a storage driver with the configured middlewares
func applyStorageMiddleware(driver storagedriver.StorageDriver, middlewares []configuration.Middleware) (storagedriver.StorageDriver, error) {
for _, mw := range middlewares {
smw, err := storagemiddleware.Get(mw.Name, mw.Options, driver)
if err != nil {
return nil, fmt.Errorf("unable to configure storage middleware (%s): %v", mw.Name, err)
}
driver = smw
}
return driver, nil
}
// uploadPurgeDefaultConfig provides a default configuration for upload
// purging to be used in the absence of configuration in the
// confifuration file
func uploadPurgeDefaultConfig() map[interface{}]interface{} {
config := map[interface{}]interface{}{}
config["enabled"] = true
config["age"] = "168h"
config["interval"] = "24h"
config["dryrun"] = false
return config
}
func badPurgeUploadConfig(reason string) {
panic(fmt.Sprintf("Unable to parse upload purge configuration: %s", reason))
}
// startUploadPurger schedules a goroutine which will periodically
// check upload directories for old files and delete them
func startUploadPurger(ctx context.Context, storageDriver storagedriver.StorageDriver, log ctxu.Logger, config map[interface{}]interface{}) {
if config["enabled"] == false {
return
}
var purgeAgeDuration time.Duration
var err error
purgeAge, ok := config["age"]
if ok {
ageStr, ok := purgeAge.(string)
if !ok {
badPurgeUploadConfig("age is not a string")
}
purgeAgeDuration, err = time.ParseDuration(ageStr)
if err != nil {
badPurgeUploadConfig(fmt.Sprintf("Cannot parse duration: %s", err.Error()))
}
} else {
badPurgeUploadConfig("age missing")
}
var intervalDuration time.Duration
interval, ok := config["interval"]
if ok {
intervalStr, ok := interval.(string)
if !ok {
badPurgeUploadConfig("interval is not a string")
}
intervalDuration, err = time.ParseDuration(intervalStr)
if err != nil {
badPurgeUploadConfig(fmt.Sprintf("Cannot parse interval: %s", err.Error()))
}
} else {
badPurgeUploadConfig("interval missing")
}
var dryRunBool bool
dryRun, ok := config["dryrun"]
if ok {
dryRunBool, ok = dryRun.(bool)
if !ok {
badPurgeUploadConfig("cannot parse dryrun")
}
} else {
badPurgeUploadConfig("dryrun missing")
}
go func() {
rand.Seed(time.Now().Unix())
jitter := time.Duration(rand.Int()%60) * time.Minute
log.Infof("Starting upload purge in %s", jitter)
time.Sleep(jitter)
for {
storage.PurgeUploads(ctx, storageDriver, time.Now().Add(-purgeAgeDuration), !dryRunBool)
log.Infof("Starting upload purge in %s", intervalDuration)
time.Sleep(intervalDuration)
}
}()
}