1 // Copyright (C) The Arvados Authors. All rights reserved.
3 // SPDX-License-Identifier: AGPL-3.0
26 type unixVolumeAdder struct {
30 // String implements flag.Value
31 func (vs *unixVolumeAdder) String() string {
35 func (vs *unixVolumeAdder) Set(path string) error {
36 if dirs := strings.Split(path, ","); len(dirs) > 1 {
37 log.Print("DEPRECATED: using comma-separated volume list.")
38 for _, dir := range dirs {
39 if err := vs.Set(dir); err != nil {
45 vs.Config.Volumes = append(vs.Config.Volumes, &UnixVolume{
47 ReadOnly: deprecated.flagReadonly,
48 Serialize: deprecated.flagSerializeIO,
54 VolumeTypes = append(VolumeTypes, func() VolumeWithExamples { return &UnixVolume{} })
56 flag.Var(&unixVolumeAdder{theConfig}, "volumes", "see Volumes configuration")
57 flag.Var(&unixVolumeAdder{theConfig}, "volume", "see Volumes configuration")
60 // Discover adds a UnixVolume for every directory named "keep" that is
61 // located at the top level of a device- or tmpfs-backed mount point
62 // other than "/". It returns the number of volumes added.
63 func (vs *unixVolumeAdder) Discover() int {
65 f, err := os.Open(ProcMounts)
67 log.Fatalf("opening %s: %s", ProcMounts, err)
69 scanner := bufio.NewScanner(f)
71 args := strings.Fields(scanner.Text())
72 if err := scanner.Err(); err != nil {
73 log.Fatalf("reading %s: %s", ProcMounts, err)
75 dev, mount := args[0], args[1]
79 if dev != "tmpfs" && !strings.HasPrefix(dev, "/dev/") {
82 keepdir := mount + "/keep"
83 if st, err := os.Stat(keepdir); err != nil || !st.IsDir() {
86 // Set the -readonly flag (but only for this volume)
87 // if the filesystem is mounted readonly.
88 flagReadonlyWas := deprecated.flagReadonly
89 for _, fsopt := range strings.Split(args[3], ",") {
91 deprecated.flagReadonly = true
98 if err := vs.Set(keepdir); err != nil {
99 log.Printf("adding %q: %s", keepdir, err)
103 deprecated.flagReadonly = flagReadonlyWas
108 // A UnixVolume stores and retrieves blocks in a local directory.
109 type UnixVolume struct {
110 Root string // path to the volume's root directory
113 DirectoryReplication int
114 StorageClasses []string
116 // something to lock during IO, typically a sync.Mutex (or nil
122 metrics *volumeMetrics
125 // DeviceID returns a globally unique ID for the volume's root
126 // directory, consisting of the filesystem's UUID and the path from
127 // filesystem root to storage directory, joined by "/". For example,
128 // the DeviceID for a local directory "/mnt/xvda1/keep" might be
129 // "fa0b6166-3b55-4994-bd3f-92f4e00a1bb0/keep".
130 func (v *UnixVolume) DeviceID() string {
131 giveup := func(f string, args ...interface{}) string {
132 log.Printf(f+"; using blank DeviceID for volume %s", append(args, v)...)
135 buf, err := exec.Command("findmnt", "--noheadings", "--target", v.Root).CombinedOutput()
137 return giveup("findmnt: %s (%q)", err, buf)
139 findmnt := strings.Fields(string(buf))
140 if len(findmnt) < 2 {
141 return giveup("could not parse findmnt output: %q", buf)
143 fsRoot, dev := findmnt[0], findmnt[1]
145 absRoot, err := filepath.Abs(v.Root)
147 return giveup("resolving relative path %q: %s", v.Root, err)
149 realRoot, err := filepath.EvalSymlinks(absRoot)
151 return giveup("resolving symlinks in %q: %s", absRoot, err)
154 // Find path from filesystem root to realRoot
156 if strings.HasPrefix(realRoot, fsRoot+"/") {
157 fsPath = realRoot[len(fsRoot):]
158 } else if fsRoot == "/" {
160 } else if fsRoot == realRoot {
163 return giveup("findmnt reports mount point %q which is not a prefix of volume root %q", fsRoot, realRoot)
166 if !strings.HasPrefix(dev, "/") {
167 return giveup("mount %q device %q is not a path", fsRoot, dev)
170 fi, err := os.Stat(dev)
172 return giveup("stat %q: %s\n", dev, err)
174 ino := fi.Sys().(*syscall.Stat_t).Ino
176 // Find a symlink in /dev/disk/by-uuid/ whose target is (i.e.,
177 // has the same inode as) the mounted device
178 udir := "/dev/disk/by-uuid"
179 d, err := os.Open(udir)
181 return giveup("opening %q: %s", udir, err)
183 uuids, err := d.Readdirnames(0)
185 return giveup("reading %q: %s", udir, err)
187 for _, uuid := range uuids {
188 link := filepath.Join(udir, uuid)
189 fi, err = os.Stat(link)
191 log.Printf("error: stat %q: %s", link, err)
194 if fi.Sys().(*syscall.Stat_t).Ino == ino {
198 return giveup("could not find entry in %q matching %q", udir, dev)
201 // Examples implements VolumeWithExamples.
202 func (*UnixVolume) Examples() []Volume {
205 Root: "/mnt/local-disk",
207 DirectoryReplication: 1,
210 Root: "/mnt/network-disk",
212 DirectoryReplication: 2,
217 // Type implements Volume
218 func (v *UnixVolume) Type() string {
222 // Start implements Volume
223 func (v *UnixVolume) Start(m *volumeMetrics) error {
225 v.locker = &sync.Mutex{}
227 if !strings.HasPrefix(v.Root, "/") {
228 return fmt.Errorf("volume root does not start with '/': %q", v.Root)
230 if v.DirectoryReplication == 0 {
231 v.DirectoryReplication = 1
233 _, err := v.os.Stat(v.Root)
235 // Set up prometheus metrics
237 v.os.stats.statsTicker.setup(m)
238 // Periodically update free/used volume space
241 v.metrics.BytesFree.Set(float64(v.Status().BytesFree))
242 v.metrics.BytesUsed.Set(float64(v.Status().BytesUsed))
243 time.Sleep(10 * time.Second)
250 // Touch sets the timestamp for the given locator to the current time
251 func (v *UnixVolume) Touch(loc string) error {
253 v.metrics.TouchOps.Inc()
255 return MethodDisabledError
257 p := v.blockPath(loc)
258 f, err := v.os.OpenFile(p, os.O_RDWR|os.O_APPEND, 0644)
263 if err := v.lock(context.TODO()); err != nil {
267 if e := v.lockfile(f); e != nil {
270 defer v.unlockfile(f)
271 ts := syscall.NsecToTimespec(time.Now().UnixNano())
272 v.os.stats.Tick(&v.os.stats.UtimesOps)
273 err = syscall.UtimesNano(p, []syscall.Timespec{ts, ts})
274 v.os.stats.TickErr(err)
278 // Mtime returns the stored timestamp for the given locator.
279 func (v *UnixVolume) Mtime(loc string) (time.Time, error) {
280 p := v.blockPath(loc)
281 fi, err := v.os.Stat(p)
283 return time.Time{}, err
285 return fi.ModTime(), nil
288 // Lock the locker (if one is in use), open the file for reading, and
289 // call the given function if and when the file is ready to read.
290 func (v *UnixVolume) getFunc(ctx context.Context, path string, fn func(io.Reader) error) error {
291 if err := v.lock(ctx); err != nil {
295 f, err := v.os.Open(path)
300 return fn(NewCountingReader(ioutil.NopCloser(f), v.os.stats.TickInBytes))
303 // stat is os.Stat() with some extra sanity checks.
304 func (v *UnixVolume) stat(path string) (os.FileInfo, error) {
305 stat, err := v.os.Stat(path)
309 } else if stat.Size() > BlockSize {
316 // Get retrieves a block, copies it to the given slice, and returns
317 // the number of bytes copied.
318 func (v *UnixVolume) Get(ctx context.Context, loc string, buf []byte) (int, error) {
320 v.metrics.GetOps.Inc()
321 return getWithPipe(ctx, loc, buf, v)
324 // ReadBlock implements BlockReader.
325 func (v *UnixVolume) ReadBlock(ctx context.Context, loc string, w io.Writer) error {
326 path := v.blockPath(loc)
327 stat, err := v.stat(path)
329 return v.translateError(err)
331 return v.getFunc(ctx, path, func(rdr io.Reader) error {
332 n, err := io.Copy(w, rdr)
333 if err == nil && n != stat.Size() {
334 err = io.ErrUnexpectedEOF
340 // Compare returns nil if Get(loc) would return the same content as
341 // expect. It is functionally equivalent to Get() followed by
342 // bytes.Compare(), but uses less memory.
343 func (v *UnixVolume) Compare(ctx context.Context, loc string, expect []byte) error {
345 v.metrics.CompareOps.Inc()
346 path := v.blockPath(loc)
347 if _, err := v.stat(path); err != nil {
348 return v.translateError(err)
350 return v.getFunc(ctx, path, func(rdr io.Reader) error {
351 return compareReaderWithBuf(ctx, rdr, expect, loc[:32])
355 // Put stores a block of data identified by the locator string
356 // "loc". It returns nil on success. If the volume is full, it
357 // returns a FullError. If the write fails due to some other error,
358 // that error is returned.
359 func (v *UnixVolume) Put(ctx context.Context, loc string, block []byte) error {
361 v.metrics.PutOps.Inc()
362 return putWithPipe(ctx, loc, block, v)
365 // WriteBlock implements BlockWriter.
366 func (v *UnixVolume) WriteBlock(ctx context.Context, loc string, rdr io.Reader) error {
368 return MethodDisabledError
373 bdir := v.blockDir(loc)
374 if err := os.MkdirAll(bdir, 0755); err != nil {
375 log.Printf("%s: could not create directory %s: %s",
380 tmpfile, tmperr := v.os.TempFile(bdir, "tmp"+loc)
382 log.Printf("ioutil.TempFile(%s, tmp%s): %s", bdir, loc, tmperr)
386 bpath := v.blockPath(loc)
388 if err := v.lock(ctx); err != nil {
392 n, err := io.Copy(tmpfile, rdr)
393 v.os.stats.TickOutBytes(uint64(n))
395 log.Printf("%s: writing to %s: %s\n", v, bpath, err)
397 v.os.Remove(tmpfile.Name())
400 if err := tmpfile.Close(); err != nil {
401 log.Printf("closing %s: %s\n", tmpfile.Name(), err)
402 v.os.Remove(tmpfile.Name())
405 if err := v.os.Rename(tmpfile.Name(), bpath); err != nil {
406 log.Printf("rename %s %s: %s\n", tmpfile.Name(), bpath, err)
407 return v.os.Remove(tmpfile.Name())
412 // Status returns a VolumeStatus struct describing the volume's
413 // current state, or nil if an error occurs.
415 func (v *UnixVolume) Status() *VolumeStatus {
416 fi, err := v.os.Stat(v.Root)
418 log.Printf("%s: os.Stat: %s\n", v, err)
421 devnum := fi.Sys().(*syscall.Stat_t).Dev
423 var fs syscall.Statfs_t
424 if err := syscall.Statfs(v.Root, &fs); err != nil {
425 log.Printf("%s: statfs: %s\n", v, err)
428 // These calculations match the way df calculates disk usage:
429 // "free" space is measured by fs.Bavail, but "used" space
430 // uses fs.Blocks - fs.Bfree.
431 free := fs.Bavail * uint64(fs.Bsize)
432 used := (fs.Blocks - fs.Bfree) * uint64(fs.Bsize)
433 return &VolumeStatus{
441 var blockDirRe = regexp.MustCompile(`^[0-9a-f]+$`)
442 var blockFileRe = regexp.MustCompile(`^[0-9a-f]{32}$`)
444 // IndexTo writes (to the given Writer) a list of blocks found on this
445 // volume which begin with the specified prefix. If the prefix is an
446 // empty string, IndexTo writes a complete list of blocks.
448 // Each block is given in the format
450 // locator+size modification-time {newline}
454 // e4df392f86be161ca6ed3773a962b8f3+67108864 1388894303
455 // e4d41e6fd68460e0e3fc18cc746959d2+67108864 1377796043
456 // e4de7a2810f5554cd39b36d8ddb132ff+67108864 1388701136
458 func (v *UnixVolume) IndexTo(prefix string, w io.Writer) error {
460 rootdir, err := v.os.Open(v.Root)
464 defer rootdir.Close()
465 v.os.stats.Tick(&v.os.stats.ReaddirOps)
467 names, err := rootdir.Readdirnames(1)
470 } else if err != nil {
473 if !strings.HasPrefix(names[0], prefix) && !strings.HasPrefix(prefix, names[0]) {
474 // prefix excludes all blocks stored in this dir
477 if !blockDirRe.MatchString(names[0]) {
480 blockdirpath := filepath.Join(v.Root, names[0])
481 blockdir, err := v.os.Open(blockdirpath)
483 log.Print("Error reading ", blockdirpath, ": ", err)
487 v.os.stats.Tick(&v.os.stats.ReaddirOps)
489 fileInfo, err := blockdir.Readdir(1)
492 } else if err != nil {
493 log.Print("Error reading ", blockdirpath, ": ", err)
497 name := fileInfo[0].Name()
498 if !strings.HasPrefix(name, prefix) {
501 if !blockFileRe.MatchString(name) {
504 _, err = fmt.Fprint(w,
506 "+", fileInfo[0].Size(),
507 " ", fileInfo[0].ModTime().UnixNano(),
510 log.Print("Error writing : ", err)
519 // Trash trashes the block data from the unix storage
520 // If TrashLifetime == 0, the block is deleted
521 // Else, the block is renamed as path/{loc}.trash.{deadline},
522 // where deadline = now + TrashLifetime
523 func (v *UnixVolume) Trash(loc string) error {
524 // Touch() must be called before calling Write() on a block. Touch()
525 // also uses lockfile(). This avoids a race condition between Write()
526 // and Trash() because either (a) the file will be trashed and Touch()
527 // will signal to the caller that the file is not present (and needs to
528 // be re-written), or (b) Touch() will update the file's timestamp and
529 // Trash() will read the correct up-to-date timestamp and choose not to
533 return MethodDisabledError
535 if err := v.lock(context.TODO()); err != nil {
539 p := v.blockPath(loc)
540 f, err := v.os.OpenFile(p, os.O_RDWR|os.O_APPEND, 0644)
545 if e := v.lockfile(f); e != nil {
548 defer v.unlockfile(f)
550 // If the block has been PUT in the last blobSignatureTTL
551 // seconds, return success without removing the block. This
552 // protects data from garbage collection until it is no longer
553 // possible for clients to retrieve the unreferenced blocks
554 // anyway (because the permission signatures have expired).
555 if fi, err := v.os.Stat(p); err != nil {
557 } else if time.Since(fi.ModTime()) < time.Duration(theConfig.BlobSignatureTTL) {
561 if theConfig.TrashLifetime == 0 {
562 return v.os.Remove(p)
564 return v.os.Rename(p, fmt.Sprintf("%v.trash.%d", p, time.Now().Add(theConfig.TrashLifetime.Duration()).Unix()))
567 // Untrash moves block from trash back into store
568 // Look for path/{loc}.trash.{deadline} in storage,
569 // and rename the first such file as path/{loc}
570 func (v *UnixVolume) Untrash(loc string) (err error) {
572 return MethodDisabledError
575 v.os.stats.Tick(&v.os.stats.ReaddirOps)
576 files, err := ioutil.ReadDir(v.blockDir(loc))
582 return os.ErrNotExist
586 prefix := fmt.Sprintf("%v.trash.", loc)
587 for _, f := range files {
588 if strings.HasPrefix(f.Name(), prefix) {
590 err = v.os.Rename(v.blockPath(f.Name()), v.blockPath(loc))
597 if foundTrash == false {
598 return os.ErrNotExist
604 // blockDir returns the fully qualified directory name for the directory
605 // where loc is (or would be) stored on this volume.
606 func (v *UnixVolume) blockDir(loc string) string {
607 return filepath.Join(v.Root, loc[0:3])
610 // blockPath returns the fully qualified pathname for the path to loc
612 func (v *UnixVolume) blockPath(loc string) string {
613 return filepath.Join(v.blockDir(loc), loc)
616 // IsFull returns true if the free space on the volume is less than
619 func (v *UnixVolume) IsFull() (isFull bool) {
620 fullSymlink := v.Root + "/full"
622 // Check if the volume has been marked as full in the last hour.
623 if link, err := os.Readlink(fullSymlink); err == nil {
624 if ts, err := strconv.Atoi(link); err == nil {
625 fulltime := time.Unix(int64(ts), 0)
626 if time.Since(fulltime).Hours() < 1.0 {
632 if avail, err := v.FreeDiskSpace(); err == nil {
633 isFull = avail < MinFreeKilobytes
635 log.Printf("%s: FreeDiskSpace: %s\n", v, err)
639 // If the volume is full, timestamp it.
641 now := fmt.Sprintf("%d", time.Now().Unix())
642 os.Symlink(now, fullSymlink)
647 // FreeDiskSpace returns the number of unused 1k blocks available on
650 func (v *UnixVolume) FreeDiskSpace() (free uint64, err error) {
651 var fs syscall.Statfs_t
652 err = syscall.Statfs(v.Root, &fs)
654 // Statfs output is not guaranteed to measure free
655 // space in terms of 1K blocks.
656 free = fs.Bavail * uint64(fs.Bsize) / 1024
661 func (v *UnixVolume) String() string {
662 return fmt.Sprintf("[UnixVolume %s]", v.Root)
665 // Writable returns false if all future Put, Mtime, and Delete calls
666 // are expected to fail.
667 func (v *UnixVolume) Writable() bool {
671 // Replication returns the number of replicas promised by the
672 // underlying device (as specified in configuration).
673 func (v *UnixVolume) Replication() int {
674 return v.DirectoryReplication
677 // GetStorageClasses implements Volume
678 func (v *UnixVolume) GetStorageClasses() []string {
679 return v.StorageClasses
682 // InternalStats returns I/O and filesystem ops counters.
683 func (v *UnixVolume) InternalStats() interface{} {
687 // lock acquires the serialize lock, if one is in use. If ctx is done
688 // before the lock is acquired, lock returns ctx.Err() instead of
689 // acquiring the lock.
690 func (v *UnixVolume) lock(ctx context.Context) error {
694 locked := make(chan struct{})
711 // unlock releases the serialize lock, if one is in use.
712 func (v *UnixVolume) unlock() {
719 // lockfile and unlockfile use flock(2) to manage kernel file locks.
720 func (v *UnixVolume) lockfile(f *os.File) error {
721 v.os.stats.Tick(&v.os.stats.FlockOps)
722 err := syscall.Flock(int(f.Fd()), syscall.LOCK_EX)
723 v.os.stats.TickErr(err)
727 func (v *UnixVolume) unlockfile(f *os.File) error {
728 err := syscall.Flock(int(f.Fd()), syscall.LOCK_UN)
729 v.os.stats.TickErr(err)
733 // Where appropriate, translate a more specific filesystem error to an
734 // error recognized by handlers, like os.ErrNotExist.
735 func (v *UnixVolume) translateError(err error) error {
738 // stat() returns a PathError if the parent directory
739 // (not just the file itself) is missing
740 return os.ErrNotExist
746 var unixTrashLocRegexp = regexp.MustCompile(`/([0-9a-f]{32})\.trash\.(\d+)$`)
748 // EmptyTrash walks hierarchy looking for {hash}.trash.*
749 // and deletes those with deadline < now.
750 func (v *UnixVolume) EmptyTrash() {
751 var bytesDeleted, bytesInTrash int64
752 var blocksDeleted, blocksInTrash int64
754 doFile := func(path string, info os.FileInfo) {
755 if info.Mode().IsDir() {
758 matches := unixTrashLocRegexp.FindStringSubmatch(path)
759 if len(matches) != 3 {
762 deadline, err := strconv.ParseInt(matches[2], 10, 64)
764 log.Printf("EmptyTrash: %v: ParseInt(%v): %v", path, matches[2], err)
767 atomic.AddInt64(&bytesInTrash, info.Size())
768 atomic.AddInt64(&blocksInTrash, 1)
769 if deadline > time.Now().Unix() {
772 err = v.os.Remove(path)
774 log.Printf("EmptyTrash: Remove %v: %v", path, err)
777 atomic.AddInt64(&bytesDeleted, info.Size())
778 atomic.AddInt64(&blocksDeleted, 1)
785 var wg sync.WaitGroup
786 todo := make(chan dirent, theConfig.EmptyTrashWorkers)
787 for i := 0; i < 1 || i < theConfig.EmptyTrashWorkers; i++ {
791 for e := range todo {
792 doFile(e.path, e.info)
797 err := filepath.Walk(v.Root, func(path string, info os.FileInfo, err error) error {
799 log.Printf("EmptyTrash: filepath.Walk: %v: %v", path, err)
802 todo <- dirent{path, info}
809 log.Printf("EmptyTrash error for %v: %v", v.String(), err)
812 log.Printf("EmptyTrash stats for %v: Deleted %v bytes in %v blocks. Remaining in trash: %v bytes in %v blocks.", v.String(), bytesDeleted, blocksDeleted, bytesInTrash-bytesDeleted, blocksInTrash-blocksDeleted)
815 type unixStats struct {
827 func (s *unixStats) TickErr(err error) {
831 s.statsTicker.TickErr(err, fmt.Sprintf("%T", err))
834 type osWithStats struct {
838 func (o *osWithStats) Open(name string) (*os.File, error) {
839 o.stats.Tick(&o.stats.OpenOps)
840 f, err := os.Open(name)
845 func (o *osWithStats) OpenFile(name string, flag int, perm os.FileMode) (*os.File, error) {
846 o.stats.Tick(&o.stats.OpenOps)
847 f, err := os.OpenFile(name, flag, perm)
852 func (o *osWithStats) Remove(path string) error {
853 o.stats.Tick(&o.stats.UnlinkOps)
854 err := os.Remove(path)
859 func (o *osWithStats) Rename(a, b string) error {
860 o.stats.Tick(&o.stats.RenameOps)
861 err := os.Rename(a, b)
866 func (o *osWithStats) Stat(path string) (os.FileInfo, error) {
867 o.stats.Tick(&o.stats.StatOps)
868 fi, err := os.Stat(path)
873 func (o *osWithStats) TempFile(dir, base string) (*os.File, error) {
874 o.stats.Tick(&o.stats.CreateOps)
875 f, err := ioutil.TempFile(dir, base)