vbatts/containerd
1
0
Fork 0
Code Pull requests Releases Activity

snapshot: clarify active and committed snapshots

Browse source 
After receiving feedback on the `snapshot.Driver` interface, it was
found that the behavior of active and committed snapshots was confusing.
We attempt to clean this up by doing the following:

1. Define the concept of "active" and "committed" snapshots and their
lifecycle relationship. Active snapshots can be created from a parent.
Committed snapshots can only be created from active snapshots.
2. Only committed snapshots can be a parent.
3. Unify the keyspace of snapshots. For common operations, such as
removal and stat, we only have a single method that works for both
active and committed snapshots. For methods that take one or the other,
the restriction is called out. `Remove` and `Delete` are consolidated
for this purpose.
4. Define the `Info` data type to include name, parent, kind and
readonly state. This allows us to collect `Exists` and `Parent` into a
single method `Stat` and simplifies the `Walk` method, eliding `Active`.
5. The `Driver` has been renamed to `Snapshotter` due to the overuse of
the term `Driver`.

Effectively, we now have snapshots that are either active or committed.

Signed-off-by: Stephen J Day <stephen.day@docker.com>
This commit is contained in:
Stephen J Day 2017-02-06 19:26:07 -08:00
parent 42a17f9391
commit ab08944aa7
No known key found for this signature in database
GPG key ID: 67B3DED84EDC823F
2 changed files with 347 additions and 302 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

302
snapshot/driver.go
View file

@ -1,302 +0,0 @@
package snapshot
import (
"io/ioutil"
"os"
"path/filepath"
"testing"
"github.com/docker/containerd"
"github.com/docker/containerd/testutil"
"github.com/stretchr/testify/assert"
)
// Driver defines the methods required to implement a snapshot driver for
// allocating, snapshotting and mounting abstract filesystems. The model works
// by building up sets of changes with parent-child relationships.
//
// These differ from the concept of the graphdriver in that the Manager has no
// knowledge of images, layers or containers. Users simply prepare and commit
// directories. We also avoid the integration between graph driver's and the
// tar format used to represent the changesets.
//
// A snapshot represents a filesystem state. Every snapshot has a parent, where
// the empty parent is represented by the empty string. A diff can be taken
// between a parent and its snapshot to generate a classic layer.
//
// For convention, we define the following terms to be used throughout this
// interface for driver implementations:
//
// `name` - refers to a forkable snapshot, typically read only
// `key` - refers to an active transaction, either a prepare or view
// `parent` - refers to the parent in relation to a name
//
// TODO(stevvooe): Update this description when things settle.
//
// Importing a Layer
//
// To import a layer, we simply have the Manager provide a list of
// mounts to be applied such that our dst will capture a changeset. We start
// out by getting a path to the layer tar file and creating a temp location to
// unpack it to:
//
// layerPath, tmpLocation := getLayerPath(), mkTmpDir() // just a path to layer tar file.
//
// We then use a Manager to prepare the temporary location as a
// snapshot point:
//
// sm := NewManager()
// mounts, err := sm.Prepare(tmpLocation, "")
// if err != nil { ... }
//
// Note that we provide "" as the parent, since we are applying the diff to an
// empty directory. We get back a list of mounts from Manager.Prepare.
// Before proceeding, we perform all these mounts:
//
// if err := MountAll(mounts); err != nil { ... }
//
// Once the mounts are performed, our temporary location is ready to capture
// a diff. In practice, this works similar to a filesystem transaction. The
// next step is to unpack the layer. We have a special function unpackLayer
// that applies the contents of the layer to target location and calculates the
// DiffID of the unpacked layer (this is a requirement for docker
// implementation):
//
// layer, err := os.Open(layerPath)
// if err != nil { ... }
// digest, err := unpackLayer(tmpLocation, layer) // unpack into layer location
// if err != nil { ... }
//
// When the above completes, we should have a filesystem the represents the
// contents of the layer. Careful implementations should verify that digest
// matches the expected DiffID. When completed, we unmount the mounts:
//
// unmount(mounts) // optional, for now
//
// Now that we've verified and unpacked our layer, we create a location to
// commit the actual diff. For this example, we are just going to use the layer
// digest, but in practice, this will probably be the ChainID:
//
// diffPath := filepath.Join("/layers", digest) // name location for the uncompressed layer digest
// if err := sm.Commit(diffPath, tmpLocation); err != nil { ... }
//
// Now, we have a layer in the Manager that can be accessed with the
// opaque diffPath provided during commit.
//
// Importing the Next Layer
//
// Making a layer depend on the above is identical to the process described
// above except that the parent is provided as diffPath when calling
// Manager.Prepare:
//
// mounts, err := sm.Prepare(tmpLocation, parentDiffPath)
//
// The diff will be captured at tmpLocation, as the layer is applied.
//
// Running a Container
//
// To run a container, we simply provide Manager.Prepare the diffPath
// of the image we want to start the container from. After mounting, the
// prepared path can be used directly as the container's filesystem:
//
// mounts, err := sm.Prepare(containerRootFS, imageDiffPath)
//
// The returned mounts can then be passed directly to the container runtime. If
// one would like to create a new image from the filesystem,
// Manager.Commit is called:
//
// if err := sm.Commit(newImageDiff, containerRootFS); err != nil { ... }
//
// Alternatively, for most container runs, Manager.Rollback will be
// called to signal Manager to abandon the changes.
type Driver interface {
// Prepare returns a set of mounts corresponding to an active snapshot
// transaction, identified by the provided transaction key.
//
// If a parent is provided, after performing the mounts, the destination
// will start with the content of the parent. Changes to the mounted
// destination will be captured in relation to the provided parent. The
// default parent, "", is an empty directory.
//
// The changes may be saved to a new snapshot by calling Commit. When one
// is done with the transaction, Remove should be called on the key.
//
// Multiple calls to Prepare or View with the same key should fail.
Prepare(key, parent string) ([]containerd.Mount, error)
// View behaves identically to Prepare except the result may not be committed
// back to the snapshot manager. View returns a readonly view on the
// parent, with the transaction tracked by the given key.
//
// This method operates identically to Prepare, except that Mounts returned
// may have the readonly flag set. Any modifications to the underlying
// filesystem will be ignored.
//
// Commit may not be called on the provided key. To collect the resources
// associated with key, Remove must be called with key as the argument.
View(key, parent string) ([]containerd.Mount, error)
// Commit captures the changes between key and its parent into a snapshot
// identified by name. The name can then be used with the driver's other
// methods to create subsequent snapshots.
//
// A snapshot will be created under name with the parent that started the
// transaction.
//
// Commit may be called multiple times on the same key. Snapshots created
// in this manner will all reference the parent used to start the
// transaction.
Commit(name, key string) error
// Mounts returns the mounts for the transaction identified by key. Can be
// called on an read-write or readonly transaction.
//
// This can be used to recover mounts after calling View or Prepare.
Mounts(key string) ([]containerd.Mount, error)
// Remove abandons the transaction identified by key. All resources
// associated with the key will be removed.
Remove(key string) error
// Parent returns the parent of snapshot identified by name.
Parent(name string) (string, error)
// Exists returns true if the snapshot with name exists.
Exists(name string) bool
// Delete the snapshot idenfitied by name.
//
// If name has children, the operation will fail.
Delete(name string) error
// TODO(stevvooe): The methods below are still in flux. We'll need to work
// out the roles of active and committed snapshots for this to become more
// clear.
// Walk the committed snapshots.
Walk(fn func(name string) error) error
// Active will call fn for each active transaction.
Active(fn func(key string) error) error
}
// DriverSuite runs a test suite on the driver given a factory function.
func DriverSuite(t *testing.T, name string, driverFn func(root string) (Driver, func(), error)) {
t.Run("Basic", makeTest(t, name, driverFn, checkDriverBasic))
}
func makeTest(t *testing.T, name string, driverFn func(root string) (Driver, func(), error), fn func(t *testing.T, driver Driver, work string)) func(t *testing.T) {
return func(t *testing.T) {
// Make two directories: a driver root and a play area for the tests:
//
// /tmp
// work/ -> passed to test functions
// root/ -> passed to driver
//
tmpDir, err := ioutil.TempDir("", "snapshot-suite-"+name+"-")
if err != nil {
t.Fatal(err)
}
defer os.RemoveAll(tmpDir)
root := filepath.Join(tmpDir, "root")
if err := os.MkdirAll(root, 0777); err != nil {
t.Fatal(err)
}
driver, cleanup, err := driverFn(root)
if err != nil {
t.Fatal(err)
}
defer cleanup()
work := filepath.Join(tmpDir, "work")
if err := os.MkdirAll(work, 0777); err != nil {
t.Fatal(err)
}
defer testutil.DumpDir(t, tmpDir)
fn(t, driver, work)
}
}
// checkDriverBasic tests the basic workflow of a snapshot driver.
func checkDriverBasic(t *testing.T, driver Driver, work string) {
preparing := filepath.Join(work, "preparing")
if err := os.MkdirAll(preparing, 0777); err != nil {
t.Fatal(err)
}
mounts, err := driver.Prepare(preparing, "")
if err != nil {
t.Fatal(err)
}
if len(mounts) < 1 {
t.Fatal("expected mounts to have entries")
}
if err := containerd.MountAll(mounts, preparing); err != nil {
t.Fatal(err)
}
defer testutil.Unmount(t, preparing)
if err := ioutil.WriteFile(filepath.Join(preparing, "foo"), []byte("foo\n"), 0777); err != nil {
t.Fatal(err)
}
if err := os.MkdirAll(filepath.Join(preparing, "a", "b", "c"), 0755); err != nil {
t.Fatal(err)
}
committed := filepath.Join(work, "committed")
if err := driver.Commit(committed, preparing); err != nil {
t.Fatal(err)
}
parent, err := driver.Parent(committed)
if err != nil {
t.Fatal(err)
}
assert.Equal(t, parent, "")
next := filepath.Join(work, "nextlayer")
if err := os.MkdirAll(next, 0777); err != nil {
t.Fatal(err)
}
mounts, err = driver.Prepare(next, committed)
if err != nil {
t.Fatal(err)
}
if err := containerd.MountAll(mounts, next); err != nil {
t.Fatal(err)
}
defer testutil.Unmount(t, next)
if err := ioutil.WriteFile(filepath.Join(next, "bar"), []byte("bar\n"), 0777); err != nil {
t.Fatal(err)
}
// also, change content of foo to bar
if err := ioutil.WriteFile(filepath.Join(next, "foo"), []byte("bar\n"), 0777); err != nil {
t.Fatal(err)
}
if err := os.RemoveAll(filepath.Join(next, "a", "b")); err != nil {
t.Log(err)
}
nextCommitted := filepath.Join(work, "committed-next")
if err := driver.Commit(nextCommitted, next); err != nil {
t.Fatal(err)
}
parent, err = driver.Parent(nextCommitted)
if err != nil {
t.Fatal(err)
}
assert.Equal(t, parent, committed)
}

347
snapshot/snapshotter.go Normal file
View file

@ -0,0 +1,347 @@
package snapshot
import (
"io/ioutil"
"os"
"path/filepath"
"testing"
"github.com/docker/containerd"
"github.com/docker/containerd/testutil"
"github.com/stretchr/testify/assert"
)
// Kind identifies the kind of snapshot.
type Kind int
// definitions of snapshot kinds
const (
KindActive Kind = iota
KindCommitted
)
// Info provides information about a particular snapshot.
type Info struct {
Name string // name or key of snapshot
Parent string // name of parent snapshot
Kind Kind // active or committed snapshot
Readonly bool // true if readonly, only valid for active
}
// Snapshotter defines the methods required to implement a snapshot snapshotter for
// allocating, snapshotting and mounting filesystem changesets. The model works
// by building up sets of changes with parent-child relationships.
//
// A snapshot represents a filesystem state. Every snapshot has a parent, where
// the empty parent is represented by the empty string. A diff can be taken
// between a parent and its snapshot to generate a classic layer.
//
// An active snapshot is created by calling `Prepare`. After mounting, changes
// can be made to the snapshot. The act of commiting creates a committed
// snapshot. The committed snapshot will get the parent of active snapshot. The
// committed snapshot can then be used as a parent. Active snapshots can never
// act as a parent.
//
// Snapshots are best understood by their lifecycle. Active snapshots are
// always created with Prepare or View. Committed snapshots are always created
// with Commit. Active snapshots never become committed snapshots and vice
// versa. All snapshots may be removed.
//
// For consistency, we define the following terms to be used throughout this
// interface for snapshotter implementations:
//
// `key` - refers to an active snapshot
// `name` - refers to a committed snapshot
// `parent` - refers to the parent in relation
//
// Most methods take various combinations of these identifiers. Typically,
// `name` and `parent` will be used in cases where a method *only* takes
// committed snapshots. `key` will be used to refer to active snapshots in most
// cases, except where noted. All variables used to access snapshots use the
// same key space. For example, an active snapshot may not share the same key
// with a committed snapshot.
//
// We cover several examples below to demonstrate the utility of a snapshot
// snapshotter.
//
// Importing a Layer
//
// To import a layer, we simply have the Manager provide a list of
// mounts to be applied such that our dst will capture a changeset. We start
// out by getting a path to the layer tar file and creating a temp location to
// unpack it to:
//
// layerPath, tmpDir := getLayerPath(), mkTmpDir() // just a path to layer tar file.
//
// We start by using a Snapshotter to Prepare a new snapshot transaction, using a
// key and descending from the empty parent "":
//
// mounts, err := snapshotter.Prepare(key, "")
// if err != nil { ... }
//
// We get back a list of mounts from Snapshotter.Prepare, with the key identifying
// the active snapshot. Mount this to the temporary location with the
// following:
//
// if err := MountAll(mounts, tmpDir); err != nil { ... }
//
// Once the mounts are performed, our temporary location is ready to capture
// a diff. In practice, this works similar to a filesystem transaction. The
// next step is to unpack the layer. We have a special function unpackLayer
// that applies the contents of the layer to target location and calculates the
// DiffID of the unpacked layer (this is a requirement for docker
// implementation):
//
// layer, err := os.Open(layerPath)
// if err != nil { ... }
// digest, err := unpackLayer(tmpLocation, layer) // unpack into layer location
// if err != nil { ... }
//
// When the above completes, we should have a filesystem the represents the
// contents of the layer. Careful implementations should verify that digest
// matches the expected DiffID. When completed, we unmount the mounts:
//
// unmount(mounts) // optional, for now
//
// Now that we've verified and unpacked our layer, we commit the active
// snapshot to a name. For this example, we are just going to use the layer
// digest, but in practice, this will probably be the ChainID:
//
// if err := snapshotter.Commit(digest.String(), key); err != nil { ... }
//
// Now, we have a layer in the Snapshotter that can be accessed with the digest
// provided during commit. Once you have committed the snapshot, the active
// snapshot can be removed with the following:
//
// snapshotter.Remove(key)
//
// Importing the Next Layer
//
// Making a layer depend on the above is identical to the process described
// above except that the parent is provided as parent when calling
// Manager.Prepare, assuming a clean tmpLocation:
//
// mounts, err := sm.Prepare(tmpLocation, parentDigest)
//
// We then mount, apply and commit, as we did above. The new snapshot will be
// based on the content of the previous one.
//
// Running a Container
//
// To run a container, we simply provide Snapshotter.Prepare the committed image
// snapshot as the parent. After mounting, the prepared path can
// be used directly as the container's filesystem:
//
// mounts, err := sm.Prepare(containerKey, imageRootFSChainID)
//
// The returned mounts can then be passed directly to the container runtime. If
// one would like to create a new image from the filesystem, Manager.Commit is
// called:
//
// if err := sm.Commit(newImageSnapshot, containerKey); err != nil { ... }
//
// Alternatively, for most container runs, Manager.Remove will be called to
// signal the Snapshotter to abandon the changes.
type Snapshotter interface {
// Stat returns the info for an active or committed snapshot by name or
// key.
//
// Should be used for parent resolution, existence checks and to discern
// the kind of snapshot.
Stat(key string) (Info, error)
// Mounts returns the mounts for the active snapshot transaction identified
// by key. Can be called on an read-write or readonly transaction. This is
// available only for active snapshots.
//
// This can be used to recover mounts after calling View or Prepare.
Mounts(key string) ([]containerd.Mount, error)
// Prepare creates an active snapshot identified by key descending from the
// provided parent. The returned mounts can be used to mount the snapshot
// to capture changes.
//
// If a parent is provided, after performing the mounts, the destination
// will start with the content of the parent. The parent must be a
// committed snapshot. Changes to the mounted destination will be captured
// in relation to the parent. The default parent, "", is an empty
// directory.
//
// The changes may be saved to a committed snapshot by calling Commit. When
// one is done with the transaction, Remove should be called on the key.
//
// Multiple calls to Prepare or View with the same key should fail.
Prepare(key, parent string) ([]containerd.Mount, error)
// View behaves identically to Prepare except the result may not be
// committed back to the snapshot snapshotter. View returns a readonly view on
// the parent, with the active snapshot being tracked by the given key.
//
// This method operates identically to Prepare, except that Mounts returned
// may have the readonly flag set. Any modifications to the underlying
// filesystem will be ignored. Implementations may perform this in a more
// efficient manner that differs from what would be attempted with
// `Prepare`.
//
// Commit may not be called on the provided key and will return an error.
// To collect the resources associated with key, Remove must be called with
// key as the argument.
View(key, parent string) ([]containerd.Mount, error)
// Commit captures the changes between key and its parent into a snapshot
// identified by name. The name can then be used with the snapshotter's other
// methods to create subsequent snapshots.
//
// A committed snapshot will be created under name with the parent of the
// active snapshot.
//
// Commit may be called multiple times on the same key. Snapshots created
// in this manner will all reference the parent used to start the
// transaction.
Commit(name, key string) error
// Remove the committed or active snapshot by the provided key.
//
// All resources associated with the key will be removed.
//
// If the snapshot is a parent of another snapshot, its children must be
// removed before proceeding.
Remove(key string) error
// Walk the committed snapshots. For each snapshot in the snapshotter, the
// function will be called.
Walk(fn func(Info) error) error
}
// SnapshotterSuite runs a test suite on the snapshotter given a factory function.
func SnapshotterSuite(t *testing.T, name string, snapshotterFn func(root string) (Snapshotter, func(), error)) {
t.Run("Basic", makeTest(t, name, snapshotterFn, checkSnapshotterBasic))
}
func makeTest(t *testing.T, name string, snapshotterFn func(root string) (Snapshotter, func(), error), fn func(t *testing.T, snapshotter Snapshotter, work string)) func(t *testing.T) {
return func(t *testing.T) {
// Make two directories: a snapshotter root and a play area for the tests:
//
// /tmp
// work/ -> passed to test functions
// root/ -> passed to snapshotter
//
tmpDir, err := ioutil.TempDir("", "snapshot-suite-"+name+"-")
if err != nil {
t.Fatal(err)
}
defer os.RemoveAll(tmpDir)
root := filepath.Join(tmpDir, "root")
if err := os.MkdirAll(root, 0777); err != nil {
t.Fatal(err)
}
snapshotter, cleanup, err := snapshotterFn(root)
if err != nil {
t.Fatal(err)
}
defer cleanup()
work := filepath.Join(tmpDir, "work")
if err := os.MkdirAll(work, 0777); err != nil {
t.Fatal(err)
}
defer testutil.DumpDir(t, tmpDir)
fn(t, snapshotter, work)
}
}
// checkSnapshotterBasic tests the basic workflow of a snapshot snapshotter.
func checkSnapshotterBasic(t *testing.T, snapshotter Snapshotter, work string) {
preparing := filepath.Join(work, "preparing")
if err := os.MkdirAll(preparing, 0777); err != nil {
t.Fatal(err)
}
mounts, err := snapshotter.Prepare(preparing, "")
if err != nil {
t.Fatal(err)
}
if len(mounts) < 1 {
t.Fatal("expected mounts to have entries")
}
if err := containerd.MountAll(mounts, preparing); err != nil {
t.Fatal(err)
}
defer testutil.Unmount(t, preparing)
if err := ioutil.WriteFile(filepath.Join(preparing, "foo"), []byte("foo\n"), 0777); err != nil {
t.Fatal(err)
}
if err := os.MkdirAll(filepath.Join(preparing, "a", "b", "c"), 0755); err != nil {
t.Fatal(err)
}
committed := filepath.Join(work, "committed")
if err := snapshotter.Commit(committed, preparing); err != nil {
t.Fatal(err)
}
si, err := snapshotter.Stat(committed)
if err != nil {
t.Fatal(err)
}
assert.Equal(t, si.Parent, "")
next := filepath.Join(work, "nextlayer")
if err := os.MkdirAll(next, 0777); err != nil {
t.Fatal(err)
}
mounts, err = snapshotter.Prepare(next, committed)
if err != nil {
t.Fatal(err)
}
if err := containerd.MountAll(mounts, next); err != nil {
t.Fatal(err)
}
defer testutil.Unmount(t, next)
if err := ioutil.WriteFile(filepath.Join(next, "bar"), []byte("bar\n"), 0777); err != nil {
t.Fatal(err)
}
// also, change content of foo to bar
if err := ioutil.WriteFile(filepath.Join(next, "foo"), []byte("bar\n"), 0777); err != nil {
t.Fatal(err)
}
if err := os.RemoveAll(filepath.Join(next, "a", "b")); err != nil {
t.Log(err)
}
nextCommitted := filepath.Join(work, "committed-next")
if err := snapshotter.Commit(nextCommitted, next); err != nil {
t.Fatal(err)
}
si2, err := snapshotter.Stat(nextCommitted)
if err != nil {
t.Fatal(err)
}
assert.Equal(t, si2.Parent, committed)
expected := map[string]Info{
si.Name: si,
si2.Name: si2,
}
walked := map[string]Info{} // walk is not ordered
assert.NoError(t, snapshotter.Walk(func(si Info) error {
walked[si.Name] = si
return nil
}))
assert.Equal(t, expected, walked)
}