// New returns a new checker which runs on repo.
func New(repo *repository.Repository) *Checker {
c := &Checker{
packs: backend.NewIDSet(),
blobs: backend.NewIDSet(),
masterIndex: repository.NewMasterIndex(),
indexes: make(map[backend.ID]*repository.Index),
repo: repo,
}
c.blobRefs.M = make(map[backend.ID]uint)
return c
}
func (cmd CmdOptimize) Execute(args []string) error {
if len(args) != 0 {
return errors.New("optimize has no arguments")
}
repo, err := cmd.global.OpenRepository()
if err != nil {
return err
}
cmd.global.Verbosef("Create exclusive lock for repository\n")
lock, err := lockRepoExclusive(repo)
defer unlockRepo(lock)
if err != nil {
return err
}
chkr := checker.New(repo)
cmd.global.Verbosef("Load indexes\n")
_, errs := chkr.LoadIndex()
if len(errs) > 0 {
for _, err := range errs {
cmd.global.Warnf("error: %v\n", err)
}
return fmt.Errorf("LoadIndex returned errors")
}
done := make(chan struct{})
errChan := make(chan error)
go chkr.Structure(errChan, done)
for err := range errChan {
if e, ok := err.(checker.TreeError); ok {
cmd.global.Warnf("error for tree %v:\n", e.ID.Str())
for _, treeErr := range e.Errors {
cmd.global.Warnf(" %v\n", treeErr)
}
} else {
cmd.global.Warnf("error: %v\n", err)
}
}
unusedBlobs := backend.NewIDSet(chkr.UnusedBlobs()...)
cmd.global.Verbosef("%d unused blobs found, repacking...\n", len(unusedBlobs))
repacker := checker.NewRepacker(repo, unusedBlobs)
err = repacker.Repack()
if err != nil {
return err
}
cmd.global.Verbosef("repacking done\n")
return nil
}
func TestIDSet(t *testing.T) {
set := backend.NewIDSet()
for i, test := range idsetTests {
seen := set.Has(test.id)
if seen != test.seen {
t.Errorf("IDSet test %v failed: wanted %v, got %v", i, test.seen, seen)
}
set.Insert(test.id)
}
}
// Packs returns all packs in this index
func (idx *Index) Packs() backend.IDSet {
idx.m.Lock()
defer idx.m.Unlock()
packs := backend.NewIDSet()
for _, entry := range idx.pack {
packs.Insert(entry.packID)
}
return packs
}
// FindBlobsForPacks returns the set of blobs contained in a pack of packs.
func FindBlobsForPacks(repo *repository.Repository, packs backend.IDSet) (backend.IDSet, error) {
blobs := backend.NewIDSet()
for packID := range packs {
for _, packedBlob := range repo.Index().ListPack(packID) {
blobs.Insert(packedBlob.ID)
}
}
return blobs, nil
}
// FindPacksForBlobs returns the set of packs that contain the blobs.
func FindPacksForBlobs(repo *repository.Repository, blobs backend.IDSet) (backend.IDSet, error) {
packs := backend.NewIDSet()
idx := repo.Index()
for id := range blobs {
blob, err := idx.Lookup(id)
if err != nil {
return nil, err
}
packs.Insert(blob.PackID)
}
return packs, nil
}
func TestIndexPacks(t *testing.T) {
idx := repository.NewIndex()
packs := backend.NewIDSet()
for i := 0; i < 20; i++ {
packID := randomID()
idx.Store(repository.PackedBlob{
Type: pack.Data,
ID: randomID(),
PackID: packID,
Offset: 0,
Length: 23,
})
packs.Insert(packID)
}
idxPacks := idx.Packs()
Assert(t, packs.Equals(idxPacks), "packs in index do not match packs added to index")
}
// NewArchiver returns a new archiver.
func NewArchiver(repo *repository.Repository) *Archiver {
arch := &Archiver{
repo: repo,
blobToken: make(chan struct{}, maxConcurrentBlobs),
knownBlobs: struct {
backend.IDSet
sync.Mutex
}{
IDSet: backend.NewIDSet(),
},
}
for i := 0; i < maxConcurrentBlobs; i++ {
arch.blobToken <- struct{}{}
}
arch.Error = archiverAbortOnAllErrors
arch.SelectFilter = archiverAllowAllFiles
return arch
}
func TestRepacker(t *testing.T) {
WithTestEnvironment(t, checkerTestData, func(repodir string) {
repo := OpenLocalRepo(t, repodir)
OK(t, repo.LoadIndex())
repo.Backend().Remove(backend.Snapshot, "c2b53c5e6a16db92fbb9aa08bd2794c58b379d8724d661ee30d20898bdfdff22")
unusedBlobs := backend.IDSet{
ParseID("5714f7274a8aa69b1692916739dc3835d09aac5395946b8ec4f58e563947199a"): struct{}{},
ParseID("08d0444e9987fa6e35ce4232b2b71473e1a8f66b2f9664cc44dc57aad3c5a63a"): struct{}{},
ParseID("356493f0b00a614d36c698591bbb2b1d801932d85328c1f508019550034549fc"): struct{}{},
ParseID("b8a6bcdddef5c0f542b4648b2ef79bc0ed4377d4109755d2fb78aff11e042663"): struct{}{},
}
chkr := checker.New(repo)
_, errs := chkr.LoadIndex()
OKs(t, errs)
errs = checkStruct(chkr)
OKs(t, errs)
list := backend.NewIDSet(chkr.UnusedBlobs()...)
if !unusedBlobs.Equals(list) {
t.Fatalf("expected unused blobs:\n %v\ngot:\n %v", unusedBlobs, list)
}
repacker := checker.NewRepacker(repo, unusedBlobs)
OK(t, repacker.Repack())
chkr = checker.New(repo)
_, errs = chkr.LoadIndex()
OKs(t, errs)
OKs(t, checkPacks(chkr))
OKs(t, checkStruct(chkr))
blobs := chkr.UnusedBlobs()
Assert(t, len(blobs) == 0,
"expected zero unused blobs, got %v", blobs)
})
}
// Packs checks that all packs referenced in the index are still available and
// there are no packs that aren't in an index. errChan is closed after all
// packs have been checked.
func (c *Checker) Packs(errChan chan<- error, done <-chan struct{}) {
defer close(errChan)
debug.Log("Checker.Packs", "checking for %d packs", len(c.packs))
seenPacks := backend.NewIDSet()
var workerWG sync.WaitGroup
IDChan := make(chan backend.ID)
for i := 0; i < defaultParallelism; i++ {
workerWG.Add(1)
go packIDTester(c.repo, IDChan, errChan, &workerWG, done)
}
for id := range c.packs {
seenPacks.Insert(id)
IDChan <- id
}
close(IDChan)
debug.Log("Checker.Packs", "waiting for %d workers to terminate", defaultParallelism)
workerWG.Wait()
debug.Log("Checker.Packs", "workers terminated")
for id := range c.repo.List(backend.Data, done) {
debug.Log("Checker.Packs", "check data blob %v", id.Str())
if !seenPacks.Has(id) {
c.orphanedPacks = append(c.orphanedPacks, id)
select {
case <-done:
return
case errChan <- PackError{ID: id, Orphaned: true, Err: errors.New("not referenced in any index")}:
}
}
}
}
}
})
}
func TestRebuildIndexAlwaysFull(t *testing.T) {
repository.IndexFull = func(*repository.Index) bool { return true }
TestRebuildIndex(t)
}
var optimizeTests = []struct {
testFilename string
snapshots backend.IDSet
}{
{
filepath.Join("..", "..", "restic", "checker", "testdata", "checker-test-repo.tar.gz"),
backend.NewIDSet(ParseID("a13c11e582b77a693dd75ab4e3a3ba96538a056594a4b9076e4cacebe6e06d43")),
},
{
filepath.Join("testdata", "old-index-repo.tar.gz"),
nil,
},
{
filepath.Join("testdata", "old-index-repo.tar.gz"),
backend.NewIDSet(
ParseID("f7d83db709977178c9d1a09e4009355e534cde1a135b8186b8b118a3fc4fcd41"),
ParseID("51d249d28815200d59e4be7b3f21a157b864dc343353df9d8e498220c2499b02"),
),
},
}
func TestCheckRestoreNoLock(t *testing.T) {
// LoadIndex loads all index files.
func (c *Checker) LoadIndex() (hints []error, errs []error) {
debug.Log("LoadIndex", "Start")
type indexRes struct {
Index *repository.Index
ID string
}
indexCh := make(chan indexRes)
worker := func(id backend.ID, done <-chan struct{}) error {
debug.Log("LoadIndex", "worker got index %v", id)
idx, err := repository.LoadIndexWithDecoder(c.repo, id, repository.DecodeIndex)
if err == repository.ErrOldIndexFormat {
debug.Log("LoadIndex", "index %v has old format", id.Str())
hints = append(hints, ErrOldIndexFormat{id})
idx, err = repository.LoadIndexWithDecoder(c.repo, id, repository.DecodeOldIndex)
}
if err != nil {
return err
}
select {
case indexCh <- indexRes{Index: idx, ID: id.String()}:
case <-done:
}
return nil
}
var perr error
go func() {
defer close(indexCh)
debug.Log("LoadIndex", "start loading indexes in parallel")
perr = repository.FilesInParallel(c.repo.Backend(), backend.Index, defaultParallelism,
repository.ParallelWorkFuncParseID(worker))
debug.Log("LoadIndex", "loading indexes finished, error: %v", perr)
}()
done := make(chan struct{})
defer close(done)
if perr != nil {
errs = append(errs, perr)
return hints, errs
}
packToIndex := make(map[backend.ID]backend.IDSet)
for res := range indexCh {
debug.Log("LoadIndex", "process index %v", res.ID)
idxID, err := backend.ParseID(res.ID)
if err != nil {
errs = append(errs, fmt.Errorf("unable to parse as index ID: %v", res.ID))
continue
}
c.indexes[idxID] = res.Index
c.masterIndex.Insert(res.Index)
debug.Log("LoadIndex", "process blobs")
cnt := 0
for blob := range res.Index.Each(done) {
c.packs.Insert(blob.PackID)
c.blobs.Insert(blob.ID)
c.blobRefs.M[blob.ID] = 0
cnt++
if _, ok := packToIndex[blob.PackID]; !ok {
packToIndex[blob.PackID] = backend.NewIDSet()
}
packToIndex[blob.PackID].Insert(idxID)
}
debug.Log("LoadIndex", "%d blobs processed", cnt)
}
debug.Log("LoadIndex", "done, error %v", perr)
debug.Log("LoadIndex", "checking for duplicate packs")
for packID := range c.packs {
debug.Log("LoadIndex", " check pack %v: contained in %d indexes", packID.Str(), len(packToIndex[packID]))
if len(packToIndex[packID]) > 1 {
hints = append(hints, ErrDuplicatePacks{
PackID: packID,
Indexes: packToIndex[packID],
})
}
}
c.repo.SetIndex(c.masterIndex)
return hints, errs
}
func (cmd CmdRebuildIndex) RebuildIndex() error {
debug.Log("RebuildIndex.RebuildIndex", "start")
done := make(chan struct{})
defer close(done)
indexIDs := backend.NewIDSet()
for id := range cmd.repo.List(backend.Index, done) {
indexIDs.Insert(id)
}
cmd.global.Printf("rebuilding index from %d indexes\n", len(indexIDs))
debug.Log("RebuildIndex.RebuildIndex", "found %v indexes", len(indexIDs))
combinedIndex := repository.NewIndex()
packsDone := backend.NewIDSet()
type Blob struct {
id backend.ID
tpe pack.BlobType
}
blobsDone := make(map[Blob]struct{})
i := 0
for indexID := range indexIDs {
cmd.global.Printf(" loading index %v\n", i)
debug.Log("RebuildIndex.RebuildIndex", "load index %v", indexID.Str())
idx, err := repository.LoadIndex(cmd.repo, indexID.String())
if err != nil {
return err
}
debug.Log("RebuildIndex.RebuildIndex", "adding blobs from index %v", indexID.Str())
for packedBlob := range idx.Each(done) {
packsDone.Insert(packedBlob.PackID)
b := Blob{
id: packedBlob.ID,
tpe: packedBlob.Type,
}
if _, ok := blobsDone[b]; ok {
continue
}
blobsDone[b] = struct{}{}
combinedIndex.Store(packedBlob)
}
combinedIndex.AddToSupersedes(indexID)
if repository.IndexFull(combinedIndex) {
combinedIndex, err = cmd.storeIndex(combinedIndex)
if err != nil {
return err
}
}
i++
}
var err error
if combinedIndex.Length() > 0 {
combinedIndex, err = cmd.storeIndex(combinedIndex)
if err != nil {
return err
}
}
cmd.global.Printf("removing %d old indexes\n", len(indexIDs))
for id := range indexIDs {
debug.Log("RebuildIndex.RebuildIndex", "remove index %v", id.Str())
err := cmd.repo.Backend().Remove(backend.Index, id.String())
if err != nil {
debug.Log("RebuildIndex.RebuildIndex", "error removing index %v: %v", id.Str(), err)
return err
}
}
cmd.global.Printf("checking for additional packs\n")
newPacks := 0
var buf []byte
for packID := range cmd.repo.List(backend.Data, done) {
if packsDone.Has(packID) {
continue
}
debug.Log("RebuildIndex.RebuildIndex", "pack %v not indexed", packID.Str())
newPacks++
var err error
h := backend.Handle{Type: backend.Data, Name: packID.String()}
buf, err = backend.LoadAll(cmd.repo.Backend(), h, buf)
if err != nil {
debug.Log("RebuildIndex.RebuildIndex", "error while loading pack %v", packID.Str())
return fmt.Errorf("error while loading pack %v: %v", packID.Str(), err)
}
hash := backend.Hash(buf)
if !hash.Equal(packID) {
debug.Log("RebuildIndex.RebuildIndex", "Pack ID does not match, want %v, got %v", packID.Str(), hash.Str())
return fmt.Errorf("Pack ID does not match, want %v, got %v", packID.Str(), hash.Str())
}
up, err := pack.NewUnpacker(cmd.repo.Key(), bytes.NewReader(buf))
if err != nil {
debug.Log("RebuildIndex.RebuildIndex", "error while unpacking pack %v", packID.Str())
return err
}
for _, blob := range up.Entries {
debug.Log("RebuildIndex.RebuildIndex", "pack %v: blob %v", packID.Str(), blob)
combinedIndex.Store(repository.PackedBlob{
Type: blob.Type,
ID: blob.ID,
PackID: packID,
Offset: blob.Offset,
Length: blob.Length,
})
}
if repository.IndexFull(combinedIndex) {
combinedIndex, err = cmd.storeIndex(combinedIndex)
if err != nil {
return err
}
}
}
if combinedIndex.Length() > 0 {
combinedIndex, err = cmd.storeIndex(combinedIndex)
if err != nil {
return err
}
}
cmd.global.Printf("added %d packs to the index\n", newPacks)
debug.Log("RebuildIndex.RebuildIndex", "done")
return nil
}