Add Store.GarbageCollect() method

This looks in the container store for existing data dirs with ids not in
the container files and removes them. It also adds an (optional) driver
method to list available layers, then uses this and compares it to the
layers json file and removes layers that are not references.

Losing track of containers and layers can potentially happen in the
case of some kind of unclean shutdown, but mainly it happens at reboot
when using transient storage mode. Such users are recommended to run
a garbage collect at boot.

Signed-off-by: Alexander Larsson <[email protected]>

containers.go

@@ -134,6 +134,9 @@ type rwContainerStore interface {
 
 	// Containers returns a slice enumerating the known containers.
 	Containers() ([]Container, error)
+
+	// Clean up unreferenced datadirs
+	GarbageCollect() error
 }
 
 type containerStore struct {
@@ -317,6 +320,39 @@ func (r *containerStore) Containers() ([]Container, error) {
 	return containers, nil
 }
 
+// This looks for datadirs in the store directory that are not referenced
+// by the json file and removes it. These can happen in the case of unclean
+// shutdowns or regular restarts in transient store mode.
+func (r *containerStore) GarbageCollect() error {
+	entries, err := os.ReadDir(r.dir)
+	if err != nil {
+		// Unexpected, don't try any GC
+		return err
+	}
+
+	for _, entry := range entries {
+		id := entry.Name()
+		// Does it look like a datadir directory?
+		if !entry.IsDir() || !nameLooksLikeID(id) {
+			continue
+		}
+
+		// Should the id be there?
+		if r.byid[id] != nil {
+			continue
+		}
+
+		// Otherwise remove datadir
+		moreErr := os.RemoveAll(filepath.Join(r.dir, id))
+		// Propagate first error
+		if moreErr != nil && err == nil {
+			err = moreErr
+		}
+	}
+
+	return err
+}
+
 func (r *containerStore) datadir(id string) string {
 	return filepath.Join(r.dir, id)
 }

drivers/aufs/aufs.go

@@ -251,6 +251,11 @@ func (a *Driver) Exists(id string) bool {
 	return true
 }
 
+// List layers (not including additional image stores)
+func (a *Driver) ListLayers() ([]string, error) {
+	return nil, graphdriver.ErrNotSupported
+}
+
 // AdditionalImageStores returns additional image stores supported by the driver
 func (a *Driver) AdditionalImageStores() []string {
 	return nil

drivers/btrfs/btrfs.go

@@ -676,6 +676,11 @@ func (d *Driver) Exists(id string) bool {
 	return err == nil
 }
 
+// List layers (not including additional image stores)
+func (d *Driver) ListLayers() ([]string, error) {
+	return nil, graphdriver.ErrNotSupported
+}
+
 // AdditionalImageStores returns additional image stores supported by the driver
 func (d *Driver) AdditionalImageStores() []string {
 	return nil

drivers/devmapper/driver.go

@@ -267,6 +267,11 @@ func (d *Driver) Exists(id string) bool {
 	return d.DeviceSet.HasDevice(id)
 }
 
+// List layers (not including additional image stores)
+func (d *Driver) ListLayers() ([]string, error) {
+	return nil, graphdriver.ErrNotSupported
+}
+
 // AdditionalImageStores returns additional image stores supported by the driver
 func (d *Driver) AdditionalImageStores() []string {
 	return nil

drivers/driver.go

@@ -109,6 +109,9 @@ type ProtoDriver interface {
 	// Exists returns whether a filesystem layer with the specified
 	// ID exists on this driver.
 	Exists(id string) bool
+	// Returns a list of layer ids that exist on this driver (does not include
+	// additional storage layers). Not supported by all backends.
+	ListLayers() ([]string, error)
 	// Status returns a set of key-value pairs which give low
 	// level diagnostic status about this driver.
 	Status() [][2]string

drivers/overlay/overlay.go

@@ -17,6 +17,7 @@ import (
 	"strings"
 	"sync"
 	"syscall"
+	"unicode"
 
 	graphdriver "github.com/containers/storage/drivers"
 	"github.com/containers/storage/drivers/overlayutils"
@@ -1697,6 +1698,40 @@ func (d *Driver) Exists(id string) bool {
 	return err == nil
 }
 
+func nameLooksLikeID(name string) bool {
+	if len(name) != 64 {
+		return false
+	}
+	for _, c := range name {
+		if !unicode.Is(unicode.ASCII_Hex_Digit, c) {
+			return false
+		}
+	}
+	return true
+}
+
+// List layers (not including additional image stores)
+func (d *Driver) ListLayers() ([]string, error) {
+	entries, err := os.ReadDir(d.home)
+	if err != nil {
+		return nil, err
+	}
+
+	layers := make([]string, 0)
+
+	for _, entry := range entries {
+		id := entry.Name()
+		// Does it look like a datadir directory?
+		if !entry.IsDir() || !nameLooksLikeID(id) {
+			continue
+		}
+
+		layers = append(layers, id)
+	}
+
+	return layers, err
+}
+
 // isParent returns if the passed in parent is the direct parent of the passed in layer
 func (d *Driver) isParent(id, parent string) bool {
 	lowers, err := d.getLowerDirs(id)

drivers/vfs/driver.go

@@ -8,6 +8,7 @@ import (
 	"runtime"
 	"strconv"
 	"strings"
+	"unicode"
 
 	graphdriver "github.com/containers/storage/drivers"
 	"github.com/containers/storage/pkg/archive"
@@ -265,6 +266,40 @@ func (d *Driver) Exists(id string) bool {
 	return err == nil
 }
 
+func nameLooksLikeID(name string) bool {
+	if len(name) != 64 {
+		return false
+	}
+	for _, c := range name {
+		if !unicode.Is(unicode.ASCII_Hex_Digit, c) {
+			return false
+		}
+	}
+	return true
+}
+
+// List layers (not including additional image stores)
+func (d *Driver) ListLayers() ([]string, error) {
+	entries, err := os.ReadDir(d.homes[0])
+	if err != nil {
+		return nil, err
+	}
+
+	layers := make([]string, 0)
+
+	for _, entry := range entries {
+		id := entry.Name()
+		// Does it look like a datadir directory?
+		if !entry.IsDir() || !nameLooksLikeID(id) {
+			continue
+		}
+
+		layers = append(layers, id)
+	}
+
+	return layers, err
+}
+
 // AdditionalImageStores returns additional image stores supported by the driver
 func (d *Driver) AdditionalImageStores() []string {
 	if len(d.homes) > 1 {

drivers/windows/windows.go

@@ -185,6 +185,11 @@ func (d *Driver) Exists(id string) bool {
 	return result
 }
 
+// List layers (not including additional image stores)
+func (d *Driver) ListLayers() ([]string, error) {
+	return nil, graphdriver.ErrNotSupported
+}
+
 // CreateFromTemplate creates a layer with the same contents and parent as another layer.
 func (d *Driver) CreateFromTemplate(id, template string, templateIDMappings *idtools.IDMappings, parent string, parentIDMappings *idtools.IDMappings, opts *graphdriver.CreateOpts, readWrite bool) error {
 	return graphdriver.NaiveCreateFromTemplate(d, id, template, templateIDMappings, parent, parentIDMappings, opts, readWrite)

drivers/zfs/zfs.go

@@ -506,6 +506,11 @@ func (d *Driver) Exists(id string) bool {
 	return d.filesystemsCache[d.zfsPath(id)]
 }
 
+// List layers (not including additional image stores)
+func (d *Driver) ListLayers() ([]string, error) {
+	return nil, graphdriver.ErrNotSupported
+}
+
 // AdditionalImageStores returns additional image stores supported by the driver
 func (d *Driver) AdditionalImageStores() []string {
 	return nil

layers.go

@@ -295,6 +295,9 @@ type rwLayerStore interface {
 	// store.
 	// This API is experimental and can be changed without bumping the major version number.
 	PutAdditionalLayer(id string, parentLayer *Layer, names []string, aLayer drivers.AdditionalLayer) (layer *Layer, err error)
+
+	// Clean up unreferenced layers
+	GarbageCollect() error
 }
 
 type layerStore struct {
@@ -512,6 +515,33 @@ func (r *layerStore) Layers() ([]Layer, error) {
 	return layers, nil
 }
 
+func (r *layerStore) GarbageCollect() error {
+	layers, err := r.driver.ListLayers()
+
+	if err != nil {
+		if errors.Is(err, drivers.ErrNotSupported) {
+			return nil
+		}
+		return err
+	}
+
+	for _, id := range layers {
+		// Is the id still referenced
+		if r.byid[id] != nil {
+			continue
+		}
+
+		// Remove layer and any related data of unreferenced id
+		if err := r.driver.Remove(id); err != nil {
+			return err
+		}
+
+		os.Remove(r.tspath(id))
+		os.RemoveAll(r.datadir(id))
+	}
+	return nil
+}
+
 func (r *layerStore) mountspath() string {
 	return filepath.Join(r.rundir, "mountpoints.json")
 }

store.go

@@ -503,6 +503,11 @@ type Store interface {
 	// Releasing AdditionalLayer handler is caller's responsibility.
 	// This API is experimental and can be changed without bumping the major version number.
 	LookupAdditionalLayer(d digest.Digest, imageref string) (AdditionalLayer, error)
+
+	// Tries to clean up remainders of previous containers or layers that are not
+	// references in the json files. These can happen in the case of unclean
+	// shutdowns or regular restarts in transient store mode.
+	GarbageCollect() error
 }
 
 // AdditionalLayer reprents a layer that is contained in the additional layer store
@@ -3385,3 +3390,20 @@ func (s *store) Free() {
 		}
 	}
 }
+
+// Tries to clean up old unreferenced container leftovers. returns the first error
+// but continues as far as it can
+func (s *store) GarbageCollect() error {
+	firstErr := s.writeToContainerStore(func(rcstore rwContainerStore) error {
+		return rcstore.GarbageCollect()
+	})
+
+	moreErr := s.writeToLayerStore(func(rlstore rwLayerStore) error {
+		return rlstore.GarbageCollect()
+	})
+	if firstErr == nil {
+		firstErr = moreErr
+	}
+
+	return firstErr
+}

utils.go

@@ -2,6 +2,7 @@ package storage
 
 import (
 	"fmt"
+	"unicode"
 
 	"github.com/containers/storage/types"
 )
@@ -72,3 +73,15 @@ func applyNameOperation(oldNames []string, opParameters []string, op updateNameO
 	}
 	return dedupeNames(result), nil
 }
+
+func nameLooksLikeID(name string) bool {
+	if len(name) != 64 {
+		return false
+	}
+	for _, c := range name {
+		if !unicode.Is(unicode.ASCII_Hex_Digit, c) {
+			return false
+		}
+	}
+	return true
+}