containerd/image.go
Wei Fu 8113758568 CRI: improve image pulling performance
Background:

With current design, the content backend uses key-lock for long-lived
write transaction. If the content reference has been marked for write
transaction, the other requestes on the same reference will fail fast with
unavailable error. Since the metadata plugin is based on boltbd which
only supports single-writer, the content backend can't block or handle
the request too long. It requires the client to handle retry by itself,
like OpenWriter - backoff retry helper. But the maximum retry interval
can be up to 2 seconds. If there are several concurrent requestes fo the
same image, the waiters maybe wakeup at the same time and there is only
one waiter can continue. A lot of waiters will get into sleep and we will
take long time to finish all the pulling jobs and be worse if the image
has many more layers, which mentioned in issue #4937.

After fetching, containerd.Pull API allows several hanlers to commit
same ChainID snapshotter but only one can be done successfully. Since
unpack tar.gz is time-consuming job, it can impact the performance on
unpacking for same ChainID snapshotter in parallel.

For instance, the Request 2 doesn't need to prepare and commit, it
should just wait for Request 1 finish, which mentioned in pull
request #6318.

```text
	Request 1	Request 2

	Prepare
	   |
	   |
	   |
	   |		Prepare
	Commit		   |
			   |
			   |
			   |
			Commit(failed on exist)
```

Both content backoff retry and unnecessary unpack impacts the performance.

Solution:

Introduced the duplicate suppression in fetch and unpack context. The
deplicate suppression uses key-mutex and single-waiter-notify to support
singleflight. The caller can use the duplicate suppression in different
PullImage handlers so that we can avoid unnecessary unpack and spin-lock
in OpenWriter.

Test Result:

Before enhancement:

```bash
➜  /tmp sudo bash testing.sh "localhost:5000/redis:latest" 20
crictl pull localhost:5000/redis:latest (x20) takes ...

real	1m6.172s
user	0m0.268s
sys	0m0.193s

docker pull localhost:5000/redis:latest (x20) takes ...

real	0m1.324s
user	0m0.441s
sys	0m0.316s

➜  /tmp sudo bash testing.sh "localhost:5000/golang:latest" 20
crictl pull localhost:5000/golang:latest (x20) takes ...

real	1m47.657s
user	0m0.284s
sys	0m0.224s

docker pull localhost:5000/golang:latest (x20) takes ...

real	0m6.381s
user	0m0.488s
sys	0m0.358s
```

With this enhancement:

```bash
➜  /tmp sudo bash testing.sh "localhost:5000/redis:latest" 20
crictl pull localhost:5000/redis:latest (x20) takes ...

real	0m1.140s
user	0m0.243s
sys	0m0.178s

docker pull localhost:5000/redis:latest (x20) takes ...

real	0m1.239s
user	0m0.463s
sys	0m0.275s

➜  /tmp sudo bash testing.sh "localhost:5000/golang:latest" 20
crictl pull localhost:5000/golang:latest (x20) takes ...

real	0m5.546s
user	0m0.217s
sys	0m0.219s

docker pull localhost:5000/golang:latest (x20) takes ...

real	0m6.090s
user	0m0.501s
sys	0m0.331s
```

Test Script:

localhost:5000/{redis|golang}:latest is equal to
docker.io/library/{redis|golang}:latest. The image is hold in local registry
service by `docker run -d -p 5000:5000 --name registry registry:2`.

```bash

image_name="${1}"
pull_times="${2:-10}"

cleanup() {
  ctr image rmi "${image_name}"
  ctr -n k8s.io image rmi "${image_name}"
  crictl rmi "${image_name}"
  docker rmi "${image_name}"
  sleep 2
}

crictl_testing() {
  for idx in $(seq 1 ${pull_times}); do
    crictl pull "${image_name}" > /dev/null 2>&1 &
  done
  wait
}

docker_testing() {
  for idx in $(seq 1 ${pull_times}); do
    docker pull "${image_name}" > /dev/null 2>&1 &
  done
  wait
}

cleanup > /dev/null 2>&1

echo 3 > /proc/sys/vm/drop_caches
sleep 3
echo "crictl pull $image_name (x${pull_times}) takes ..."
time crictl_testing
echo

echo 3 > /proc/sys/vm/drop_caches
sleep 3
echo "docker pull $image_name (x${pull_times}) takes ..."
time docker_testing
```

Fixes: #4937
Close: #4985
Close: #6318

Signed-off-by: Wei Fu <fuweid89@gmail.com>
2022-04-06 07:14:18 +08:00

470 lines
13 KiB
Go

/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package containerd
import (
"context"
"encoding/json"
"errors"
"fmt"
"strings"
"sync/atomic"
"github.com/containerd/containerd/content"
"github.com/containerd/containerd/diff"
"github.com/containerd/containerd/errdefs"
"github.com/containerd/containerd/images"
"github.com/containerd/containerd/pkg/kmutex"
"github.com/containerd/containerd/platforms"
"github.com/containerd/containerd/rootfs"
"github.com/containerd/containerd/snapshots"
"github.com/opencontainers/go-digest"
"github.com/opencontainers/image-spec/identity"
ocispec "github.com/opencontainers/image-spec/specs-go/v1"
"golang.org/x/sync/semaphore"
)
// Image describes an image used by containers
type Image interface {
// Name of the image
Name() string
// Target descriptor for the image content
Target() ocispec.Descriptor
// Labels of the image
Labels() map[string]string
// Unpack unpacks the image's content into a snapshot
Unpack(context.Context, string, ...UnpackOpt) error
// RootFS returns the unpacked diffids that make up images rootfs.
RootFS(ctx context.Context) ([]digest.Digest, error)
// Size returns the total size of the image's packed resources.
Size(ctx context.Context) (int64, error)
// Usage returns a usage calculation for the image.
Usage(context.Context, ...UsageOpt) (int64, error)
// Config descriptor for the image.
Config(ctx context.Context) (ocispec.Descriptor, error)
// IsUnpacked returns whether or not an image is unpacked.
IsUnpacked(context.Context, string) (bool, error)
// ContentStore provides a content store which contains image blob data
ContentStore() content.Store
// Metadata returns the underlying image metadata
Metadata() images.Image
// Platform returns the platform match comparer. Can be nil.
Platform() platforms.MatchComparer
}
type usageOptions struct {
manifestLimit *int
manifestOnly bool
snapshots bool
}
// UsageOpt is used to configure the usage calculation
type UsageOpt func(*usageOptions) error
// WithUsageManifestLimit sets the limit to the number of manifests which will
// be walked for usage. Setting this value to 0 will require all manifests to
// be walked, returning ErrNotFound if manifests are missing.
// NOTE: By default all manifests which exist will be walked
// and any non-existent manifests and their subobjects will be ignored.
func WithUsageManifestLimit(i int) UsageOpt {
// If 0 then don't filter any manifests
// By default limits to current platform
return func(o *usageOptions) error {
o.manifestLimit = &i
return nil
}
}
// WithSnapshotUsage will check for referenced snapshots from the image objects
// and include the snapshot size in the total usage.
func WithSnapshotUsage() UsageOpt {
return func(o *usageOptions) error {
o.snapshots = true
return nil
}
}
// WithManifestUsage is used to get the usage for an image based on what is
// reported by the manifests rather than what exists in the content store.
// NOTE: This function is best used with the manifest limit set to get a
// consistent value, otherwise non-existent manifests will be excluded.
func WithManifestUsage() UsageOpt {
return func(o *usageOptions) error {
o.manifestOnly = true
return nil
}
}
var _ = (Image)(&image{})
// NewImage returns a client image object from the metadata image
func NewImage(client *Client, i images.Image) Image {
return &image{
client: client,
i: i,
platform: client.platform,
}
}
// NewImageWithPlatform returns a client image object from the metadata image
func NewImageWithPlatform(client *Client, i images.Image, platform platforms.MatchComparer) Image {
return &image{
client: client,
i: i,
platform: platform,
}
}
type image struct {
client *Client
i images.Image
platform platforms.MatchComparer
}
func (i *image) Metadata() images.Image {
return i.i
}
func (i *image) Name() string {
return i.i.Name
}
func (i *image) Target() ocispec.Descriptor {
return i.i.Target
}
func (i *image) Labels() map[string]string {
return i.i.Labels
}
func (i *image) RootFS(ctx context.Context) ([]digest.Digest, error) {
provider := i.client.ContentStore()
return i.i.RootFS(ctx, provider, i.platform)
}
func (i *image) Size(ctx context.Context) (int64, error) {
return i.Usage(ctx, WithUsageManifestLimit(1), WithManifestUsage())
}
func (i *image) Usage(ctx context.Context, opts ...UsageOpt) (int64, error) {
var config usageOptions
for _, opt := range opts {
if err := opt(&config); err != nil {
return 0, err
}
}
var (
provider = i.client.ContentStore()
handler = images.ChildrenHandler(provider)
size int64
mustExist bool
)
if config.manifestLimit != nil {
handler = images.LimitManifests(handler, i.platform, *config.manifestLimit)
mustExist = true
}
var wh images.HandlerFunc = func(ctx context.Context, desc ocispec.Descriptor) ([]ocispec.Descriptor, error) {
var usage int64
children, err := handler(ctx, desc)
if err != nil {
if !errdefs.IsNotFound(err) || mustExist {
return nil, err
}
if !config.manifestOnly {
// Do not count size of non-existent objects
desc.Size = 0
}
} else if config.snapshots || !config.manifestOnly {
info, err := provider.Info(ctx, desc.Digest)
if err != nil {
if !errdefs.IsNotFound(err) {
return nil, err
}
if !config.manifestOnly {
// Do not count size of non-existent objects
desc.Size = 0
}
} else if info.Size > desc.Size {
// Count actual usage, Size may be unset or -1
desc.Size = info.Size
}
if config.snapshots {
for k, v := range info.Labels {
const prefix = "containerd.io/gc.ref.snapshot."
if !strings.HasPrefix(k, prefix) {
continue
}
sn := i.client.SnapshotService(k[len(prefix):])
if sn == nil {
continue
}
u, err := sn.Usage(ctx, v)
if err != nil {
if !errdefs.IsNotFound(err) && !errdefs.IsInvalidArgument(err) {
return nil, err
}
} else {
usage += u.Size
}
}
}
}
// Ignore unknown sizes. Generally unknown sizes should
// never be set in manifests, however, the usage
// calculation does not need to enforce this.
if desc.Size >= 0 {
usage += desc.Size
}
atomic.AddInt64(&size, usage)
return children, nil
}
l := semaphore.NewWeighted(3)
if err := images.Dispatch(ctx, wh, l, i.i.Target); err != nil {
return 0, err
}
return size, nil
}
func (i *image) Config(ctx context.Context) (ocispec.Descriptor, error) {
provider := i.client.ContentStore()
return i.i.Config(ctx, provider, i.platform)
}
func (i *image) IsUnpacked(ctx context.Context, snapshotterName string) (bool, error) {
sn, err := i.client.getSnapshotter(ctx, snapshotterName)
if err != nil {
return false, err
}
cs := i.client.ContentStore()
diffs, err := i.i.RootFS(ctx, cs, i.platform)
if err != nil {
return false, err
}
chainID := identity.ChainID(diffs)
_, err = sn.Stat(ctx, chainID.String())
if err == nil {
return true, nil
} else if !errdefs.IsNotFound(err) {
return false, err
}
return false, nil
}
// UnpackConfig provides configuration for the unpack of an image
type UnpackConfig struct {
// ApplyOpts for applying a diff to a snapshotter
ApplyOpts []diff.ApplyOpt
// SnapshotOpts for configuring a snapshotter
SnapshotOpts []snapshots.Opt
// CheckPlatformSupported is whether to validate that a snapshotter
// supports an image's platform before unpacking
CheckPlatformSupported bool
// DuplicationSuppressor is used to make sure that there is only one
// in-flight fetch request or unpack handler for a given descriptor's
// digest or chain ID.
DuplicationSuppressor kmutex.KeyedLocker
}
// UnpackOpt provides configuration for unpack
type UnpackOpt func(context.Context, *UnpackConfig) error
// WithSnapshotterPlatformCheck sets `CheckPlatformSupported` on the UnpackConfig
func WithSnapshotterPlatformCheck() UnpackOpt {
return func(ctx context.Context, uc *UnpackConfig) error {
uc.CheckPlatformSupported = true
return nil
}
}
// WithUnpackDuplicationSuppressor sets `DuplicationSuppressor` on the UnpackConfig.
func WithUnpackDuplicationSuppressor(suppressor kmutex.KeyedLocker) UnpackOpt {
return func(ctx context.Context, uc *UnpackConfig) error {
uc.DuplicationSuppressor = suppressor
return nil
}
}
func (i *image) Unpack(ctx context.Context, snapshotterName string, opts ...UnpackOpt) error {
ctx, done, err := i.client.WithLease(ctx)
if err != nil {
return err
}
defer done(ctx)
var config UnpackConfig
for _, o := range opts {
if err := o(ctx, &config); err != nil {
return err
}
}
manifest, err := i.getManifest(ctx, i.platform)
if err != nil {
return err
}
layers, err := i.getLayers(ctx, i.platform, manifest)
if err != nil {
return err
}
var (
a = i.client.DiffService()
cs = i.client.ContentStore()
chain []digest.Digest
unpacked bool
)
snapshotterName, err = i.client.resolveSnapshotterName(ctx, snapshotterName)
if err != nil {
return err
}
sn, err := i.client.getSnapshotter(ctx, snapshotterName)
if err != nil {
return err
}
if config.CheckPlatformSupported {
if err := i.checkSnapshotterSupport(ctx, snapshotterName, manifest); err != nil {
return err
}
}
for _, layer := range layers {
unpacked, err = rootfs.ApplyLayerWithOpts(ctx, layer, chain, sn, a, config.SnapshotOpts, config.ApplyOpts)
if err != nil {
return err
}
if unpacked {
// Set the uncompressed label after the uncompressed
// digest has been verified through apply.
cinfo := content.Info{
Digest: layer.Blob.Digest,
Labels: map[string]string{
"containerd.io/uncompressed": layer.Diff.Digest.String(),
},
}
if _, err := cs.Update(ctx, cinfo, "labels.containerd.io/uncompressed"); err != nil {
return err
}
}
chain = append(chain, layer.Diff.Digest)
}
desc, err := i.i.Config(ctx, cs, i.platform)
if err != nil {
return err
}
rootfs := identity.ChainID(chain).String()
cinfo := content.Info{
Digest: desc.Digest,
Labels: map[string]string{
fmt.Sprintf("containerd.io/gc.ref.snapshot.%s", snapshotterName): rootfs,
},
}
_, err = cs.Update(ctx, cinfo, fmt.Sprintf("labels.containerd.io/gc.ref.snapshot.%s", snapshotterName))
return err
}
func (i *image) getManifest(ctx context.Context, platform platforms.MatchComparer) (ocispec.Manifest, error) {
cs := i.ContentStore()
manifest, err := images.Manifest(ctx, cs, i.i.Target, platform)
if err != nil {
return ocispec.Manifest{}, err
}
return manifest, nil
}
func (i *image) getLayers(ctx context.Context, platform platforms.MatchComparer, manifest ocispec.Manifest) ([]rootfs.Layer, error) {
cs := i.ContentStore()
diffIDs, err := i.i.RootFS(ctx, cs, platform)
if err != nil {
return nil, fmt.Errorf("failed to resolve rootfs: %w", err)
}
if len(diffIDs) != len(manifest.Layers) {
return nil, errors.New("mismatched image rootfs and manifest layers")
}
layers := make([]rootfs.Layer, len(diffIDs))
for i := range diffIDs {
layers[i].Diff = ocispec.Descriptor{
// TODO: derive media type from compressed type
MediaType: ocispec.MediaTypeImageLayer,
Digest: diffIDs[i],
}
layers[i].Blob = manifest.Layers[i]
}
return layers, nil
}
func (i *image) getManifestPlatform(ctx context.Context, manifest ocispec.Manifest) (ocispec.Platform, error) {
cs := i.ContentStore()
p, err := content.ReadBlob(ctx, cs, manifest.Config)
if err != nil {
return ocispec.Platform{}, err
}
var image ocispec.Image
if err := json.Unmarshal(p, &image); err != nil {
return ocispec.Platform{}, err
}
return platforms.Normalize(ocispec.Platform{OS: image.OS, Architecture: image.Architecture}), nil
}
func (i *image) checkSnapshotterSupport(ctx context.Context, snapshotterName string, manifest ocispec.Manifest) error {
snapshotterPlatformMatcher, err := i.client.GetSnapshotterSupportedPlatforms(ctx, snapshotterName)
if err != nil {
return err
}
manifestPlatform, err := i.getManifestPlatform(ctx, manifest)
if err != nil {
return err
}
if snapshotterPlatformMatcher.Match(manifestPlatform) {
return nil
}
return fmt.Errorf("snapshotter %s does not support platform %s for image %s", snapshotterName, manifestPlatform, manifest.Config.Digest)
}
func (i *image) ContentStore() content.Store {
return i.client.ContentStore()
}
func (i *image) Platform() platforms.MatchComparer {
return i.platform
}