1 // Copyright (C) The Arvados Authors. All rights reserved.
3 // SPDX-License-Identifier: AGPL-3.0
23 "git.curoverse.com/arvados.git/sdk/go/arvados"
24 "github.com/AdRoll/goamz/aws"
25 "github.com/AdRoll/goamz/s3"
26 log "github.com/Sirupsen/logrus"
30 s3DefaultReadTimeout = arvados.Duration(10 * time.Minute)
31 s3DefaultConnectTimeout = arvados.Duration(time.Minute)
35 // ErrS3TrashDisabled is returned by Trash if that operation
36 // is impossible with the current config.
37 ErrS3TrashDisabled = fmt.Errorf("trash function is disabled because -trash-lifetime=0 and -s3-unsafe-delete=false")
39 s3AccessKeyFile string
40 s3SecretKeyFile string
45 s3RaceWindow time.Duration
53 maxClockSkew = 600 * time.Second
54 nearlyRFC1123 = "Mon, 2 Jan 2006 15:04:05 GMT"
57 type s3VolumeAdder struct {
61 // String implements flag.Value
62 func (s *s3VolumeAdder) String() string {
66 func (s *s3VolumeAdder) Set(bucketName string) error {
68 return fmt.Errorf("no container name given")
70 if s3AccessKeyFile == "" || s3SecretKeyFile == "" {
71 return fmt.Errorf("-s3-access-key-file and -s3-secret-key-file arguments must given before -s3-bucket-volume")
73 if deprecated.flagSerializeIO {
74 log.Print("Notice: -serialize is not supported by s3-bucket volumes.")
76 s.Config.Volumes = append(s.Config.Volumes, &S3Volume{
78 AccessKeyFile: s3AccessKeyFile,
79 SecretKeyFile: s3SecretKeyFile,
82 RaceWindow: arvados.Duration(s3RaceWindow),
83 S3Replication: s3Replication,
84 UnsafeDelete: s3UnsafeDelete,
85 ReadOnly: deprecated.flagReadonly,
91 func s3regions() (okList []string) {
92 for r := range aws.Regions {
93 okList = append(okList, r)
99 VolumeTypes = append(VolumeTypes, func() VolumeWithExamples { return &S3Volume{} })
101 flag.Var(&s3VolumeAdder{theConfig},
103 "Use the given bucket as a storage volume. Can be given multiple times.")
108 fmt.Sprintf("AWS region used for subsequent -s3-bucket-volume arguments. Allowed values are %+q.", s3regions()))
113 "Endpoint URL used for subsequent -s3-bucket-volume arguments. If blank, use the AWS endpoint corresponding to the -s3-region argument. For Google Storage, use \"https://storage.googleapis.com\".")
116 "s3-access-key-file",
118 "`File` containing the access key used for subsequent -s3-bucket-volume arguments.")
121 "s3-secret-key-file",
123 "`File` containing the secret key used for subsequent -s3-bucket-volume arguments.")
128 "Maximum eventual consistency latency for subsequent -s3-bucket-volume arguments.")
133 "Replication level reported to clients for subsequent -s3-bucket-volume arguments.")
138 "EXPERIMENTAL. Enable deletion (garbage collection) even when trash lifetime is zero, even though there are known race conditions that can cause data loss.")
141 // S3Volume implements Volume using an S3 bucket.
142 type S3Volume struct {
148 LocationConstraint bool
151 ConnectTimeout arvados.Duration
152 ReadTimeout arvados.Duration
153 RaceWindow arvados.Duration
162 // Examples implements VolumeWithExamples.
163 func (*S3Volume) Examples() []Volume {
166 AccessKeyFile: "/etc/aws_s3_access_key.txt",
167 SecretKeyFile: "/etc/aws_s3_secret_key.txt",
170 Bucket: "example-bucket-name",
173 RaceWindow: arvados.Duration(24 * time.Hour),
174 ConnectTimeout: arvados.Duration(time.Minute),
175 ReadTimeout: arvados.Duration(5 * time.Minute),
178 AccessKeyFile: "/etc/gce_s3_access_key.txt",
179 SecretKeyFile: "/etc/gce_s3_secret_key.txt",
180 Endpoint: "https://storage.googleapis.com",
182 Bucket: "example-bucket-name",
185 RaceWindow: arvados.Duration(24 * time.Hour),
186 ConnectTimeout: arvados.Duration(time.Minute),
187 ReadTimeout: arvados.Duration(5 * time.Minute),
192 // Type implements Volume.
193 func (*S3Volume) Type() string {
197 // Start populates private fields and verifies the configuration is
199 func (v *S3Volume) Start() error {
200 region, ok := aws.Regions[v.Region]
201 if v.Endpoint == "" {
203 return fmt.Errorf("unrecognized region %+q; try specifying -s3-endpoint instead", v.Region)
206 return fmt.Errorf("refusing to use AWS region name %+q with endpoint %+q; "+
207 "specify empty endpoint (\"-s3-endpoint=\") or use a different region name", v.Region, v.Endpoint)
211 S3Endpoint: v.Endpoint,
212 S3LocationConstraint: v.LocationConstraint,
218 auth.AccessKey, err = readKeyFromFile(v.AccessKeyFile)
222 auth.SecretKey, err = readKeyFromFile(v.SecretKeyFile)
227 // Zero timeouts mean "wait forever", which is a bad
228 // default. Default to long timeouts instead.
229 if v.ConnectTimeout == 0 {
230 v.ConnectTimeout = s3DefaultConnectTimeout
232 if v.ReadTimeout == 0 {
233 v.ReadTimeout = s3DefaultReadTimeout
236 client := s3.New(auth, region)
237 client.ConnectTimeout = time.Duration(v.ConnectTimeout)
238 client.ReadTimeout = time.Duration(v.ReadTimeout)
239 v.bucket = &s3bucket{
248 // DeviceID returns a globally unique ID for the storage bucket.
249 func (v *S3Volume) DeviceID() string {
250 return "s3://" + v.Endpoint + "/" + v.Bucket
253 func (v *S3Volume) getReaderWithContext(ctx context.Context, loc string) (rdr io.ReadCloser, err error) {
254 ready := make(chan bool)
256 rdr, err = v.getReader(loc)
263 theConfig.debugLogf("s3: abandoning getReader(): %s", ctx.Err())
270 return nil, ctx.Err()
274 // getReader wraps (Bucket)GetReader.
276 // In situations where (Bucket)GetReader would fail because the block
277 // disappeared in a Trash race, getReader calls fixRace to recover the
278 // data, and tries again.
279 func (v *S3Volume) getReader(loc string) (rdr io.ReadCloser, err error) {
280 rdr, err = v.bucket.GetReader(loc)
281 err = v.translateError(err)
282 if err == nil || !os.IsNotExist(err) {
286 _, err = v.bucket.Head("recent/"+loc, nil)
287 err = v.translateError(err)
289 // If we can't read recent/X, there's no point in
290 // trying fixRace. Give up.
298 rdr, err = v.bucket.GetReader(loc)
300 log.Printf("warning: reading %s after successful fixRace: %s", loc, err)
301 err = v.translateError(err)
306 // Get a block: copy the block data into buf, and return the number of
308 func (v *S3Volume) Get(ctx context.Context, loc string, buf []byte) (int, error) {
309 rdr, err := v.getReaderWithContext(ctx, loc)
315 ready := make(chan bool)
320 n, err = io.ReadFull(rdr, buf)
323 case nil, io.EOF, io.ErrUnexpectedEOF:
326 err = v.translateError(err)
331 theConfig.debugLogf("s3: interrupting ReadFull() with Close() because %s", ctx.Err())
333 // Must wait for ReadFull to return, to ensure it
334 // doesn't write to buf after we return.
335 theConfig.debugLogf("s3: waiting for ReadFull() to fail")
343 // Compare the given data with the stored data.
344 func (v *S3Volume) Compare(ctx context.Context, loc string, expect []byte) error {
345 errChan := make(chan error, 1)
347 _, err := v.bucket.Head("recent/"+loc, nil)
354 case err = <-errChan:
357 // Checking for "loc" itself here would interfere with
358 // future GET requests.
360 // On AWS, if X doesn't exist, a HEAD or GET request
361 // for X causes X's non-existence to be cached. Thus,
362 // if we test for X, then create X and return a
363 // signature to our client, the client might still get
364 // 404 from all keepstores when trying to read it.
366 // To avoid this, we avoid doing HEAD X or GET X until
367 // we know X has been written.
369 // Note that X might exist even though recent/X
370 // doesn't: for example, the response to HEAD recent/X
371 // might itself come from a stale cache. In such
372 // cases, we will return a false negative and
373 // PutHandler might needlessly create another replica
374 // on a different volume. That's not ideal, but it's
375 // better than passing the eventually-consistent
376 // problem on to our clients.
377 return v.translateError(err)
379 rdr, err := v.getReaderWithContext(ctx, loc)
384 return v.translateError(compareReaderWithBuf(ctx, rdr, expect, loc[:32]))
387 // Put writes a block.
388 func (v *S3Volume) Put(ctx context.Context, loc string, block []byte) error {
390 return MethodDisabledError
395 md5, err := hex.DecodeString(loc)
399 opts.ContentMD5 = base64.StdEncoding.EncodeToString(md5)
402 // Send the block data through a pipe, so that (if we need to)
403 // we can close the pipe early and abandon our PutReader()
404 // goroutine, without worrying about PutReader() accessing our
405 // block buffer after we release it.
406 bufr, bufw := io.Pipe()
408 io.Copy(bufw, bytes.NewReader(block))
413 ready := make(chan bool)
416 if ctx.Err() != nil {
417 theConfig.debugLogf("%s: abandoned PutReader goroutine finished with err: %s", v, err)
421 err = v.bucket.PutReader(loc, bufr, int64(size), "application/octet-stream", s3ACL, opts)
425 err = v.bucket.PutReader("recent/"+loc, nil, 0, "application/octet-stream", s3ACL, s3.Options{})
429 theConfig.debugLogf("%s: taking PutReader's input away: %s", v, ctx.Err())
430 // Our pipe might be stuck in Write(), waiting for
431 // io.Copy() to read. If so, un-stick it. This means
432 // PutReader will get corrupt data, but that's OK: the
433 // size and MD5 won't match, so the write will fail.
434 go io.Copy(ioutil.Discard, bufr)
435 // CloseWithError() will return once pending I/O is done.
436 bufw.CloseWithError(ctx.Err())
437 theConfig.debugLogf("%s: abandoning PutReader goroutine", v)
440 return v.translateError(err)
444 // Touch sets the timestamp for the given locator to the current time.
445 func (v *S3Volume) Touch(loc string) error {
447 return MethodDisabledError
449 _, err := v.bucket.Head(loc, nil)
450 err = v.translateError(err)
451 if os.IsNotExist(err) && v.fixRace(loc) {
452 // The data object got trashed in a race, but fixRace
454 } else if err != nil {
457 err = v.bucket.PutReader("recent/"+loc, nil, 0, "application/octet-stream", s3ACL, s3.Options{})
458 return v.translateError(err)
461 // Mtime returns the stored timestamp for the given locator.
462 func (v *S3Volume) Mtime(loc string) (time.Time, error) {
463 _, err := v.bucket.Head(loc, nil)
465 return zeroTime, v.translateError(err)
467 resp, err := v.bucket.Head("recent/"+loc, nil)
468 err = v.translateError(err)
469 if os.IsNotExist(err) {
470 // The data object X exists, but recent/X is missing.
471 err = v.bucket.PutReader("recent/"+loc, nil, 0, "application/octet-stream", s3ACL, s3.Options{})
473 log.Printf("error: creating %q: %s", "recent/"+loc, err)
474 return zeroTime, v.translateError(err)
476 log.Printf("info: created %q to migrate existing block to new storage scheme", "recent/"+loc)
477 resp, err = v.bucket.Head("recent/"+loc, nil)
479 log.Printf("error: created %q but HEAD failed: %s", "recent/"+loc, err)
480 return zeroTime, v.translateError(err)
482 } else if err != nil {
483 // HEAD recent/X failed for some other reason.
486 return v.lastModified(resp)
489 // IndexTo writes a complete list of locators with the given prefix
490 // for which Get() can retrieve data.
491 func (v *S3Volume) IndexTo(prefix string, writer io.Writer) error {
492 // Use a merge sort to find matching sets of X and recent/X.
494 Bucket: v.bucket.Bucket,
496 PageSize: v.IndexPageSize,
499 Bucket: v.bucket.Bucket,
500 Prefix: "recent/" + prefix,
501 PageSize: v.IndexPageSize,
503 v.bucket.stats.Tick(&v.bucket.stats.Ops, &v.bucket.stats.ListOps)
504 v.bucket.stats.Tick(&v.bucket.stats.Ops, &v.bucket.stats.ListOps)
505 for data, recent := dataL.First(), recentL.First(); data != nil; data = dataL.Next() {
506 v.bucket.stats.Tick(&v.bucket.stats.Ops, &v.bucket.stats.ListOps)
508 // Conveniently, "recent/*" and "trash/*" are
509 // lexically greater than all hex-encoded data
510 // hashes, so stopping here avoids iterating
511 // over all of them needlessly with dataL.
514 if !v.isKeepBlock(data.Key) {
518 // stamp is the list entry we should use to report the
519 // last-modified time for this data block: it will be
520 // the recent/X entry if one exists, otherwise the
521 // entry for the data block itself.
524 // Advance to the corresponding recent/X marker, if any
526 if cmp := strings.Compare(recent.Key[7:], data.Key); cmp < 0 {
527 recent = recentL.Next()
528 v.bucket.stats.Tick(&v.bucket.stats.Ops, &v.bucket.stats.ListOps)
532 recent = recentL.Next()
533 v.bucket.stats.Tick(&v.bucket.stats.Ops, &v.bucket.stats.ListOps)
536 // recent/X marker is missing: we'll
537 // use the timestamp on the data
542 t, err := time.Parse(time.RFC3339, stamp.LastModified)
546 fmt.Fprintf(writer, "%s+%d %d\n", data.Key, data.Size, t.UnixNano())
551 // Trash a Keep block.
552 func (v *S3Volume) Trash(loc string) error {
554 return MethodDisabledError
556 if t, err := v.Mtime(loc); err != nil {
558 } else if time.Since(t) < theConfig.BlobSignatureTTL.Duration() {
561 if theConfig.TrashLifetime == 0 {
563 return ErrS3TrashDisabled
565 return v.translateError(v.bucket.Del(loc))
567 err := v.checkRaceWindow(loc)
571 err = v.safeCopy("trash/"+loc, loc)
575 return v.translateError(v.bucket.Del(loc))
578 // checkRaceWindow returns a non-nil error if trash/loc is, or might
579 // be, in the race window (i.e., it's not safe to trash loc).
580 func (v *S3Volume) checkRaceWindow(loc string) error {
581 resp, err := v.bucket.Head("trash/"+loc, nil)
582 err = v.translateError(err)
583 if os.IsNotExist(err) {
584 // OK, trash/X doesn't exist so we're not in the race
587 } else if err != nil {
588 // Error looking up trash/X. We don't know whether
589 // we're in the race window
592 t, err := v.lastModified(resp)
594 // Can't parse timestamp
597 safeWindow := t.Add(theConfig.TrashLifetime.Duration()).Sub(time.Now().Add(time.Duration(v.RaceWindow)))
599 // We can't count on "touch trash/X" to prolong
600 // trash/X's lifetime. The new timestamp might not
601 // become visible until now+raceWindow, and EmptyTrash
602 // is allowed to delete trash/X before then.
603 return fmt.Errorf("same block is already in trash, and safe window ended %s ago", -safeWindow)
605 // trash/X exists, but it won't be eligible for deletion until
606 // after now+raceWindow, so it's safe to overwrite it.
610 // safeCopy calls PutCopy, and checks the response to make sure the
611 // copy succeeded and updated the timestamp on the destination object
612 // (PutCopy returns 200 OK if the request was received, even if the
614 func (v *S3Volume) safeCopy(dst, src string) error {
615 resp, err := v.bucket.PutCopy(dst, s3ACL, s3.CopyOptions{
616 ContentType: "application/octet-stream",
617 MetadataDirective: "REPLACE",
618 }, v.bucket.Name+"/"+src)
619 err = v.translateError(err)
623 if t, err := time.Parse(time.RFC3339Nano, resp.LastModified); err != nil {
624 return fmt.Errorf("PutCopy succeeded but did not return a timestamp: %q: %s", resp.LastModified, err)
625 } else if time.Now().Sub(t) > maxClockSkew {
626 return fmt.Errorf("PutCopy succeeded but returned an old timestamp: %q: %s", resp.LastModified, t)
631 // Get the LastModified header from resp, and parse it as RFC1123 or
632 // -- if it isn't valid RFC1123 -- as Amazon's variant of RFC1123.
633 func (v *S3Volume) lastModified(resp *http.Response) (t time.Time, err error) {
634 s := resp.Header.Get("Last-Modified")
635 t, err = time.Parse(time.RFC1123, s)
636 if err != nil && s != "" {
637 // AWS example is "Sun, 1 Jan 2006 12:00:00 GMT",
638 // which isn't quite "Sun, 01 Jan 2006 12:00:00 GMT"
639 // as required by HTTP spec. If it's not a valid HTTP
640 // header value, it's probably AWS (or s3test) giving
641 // us a nearly-RFC1123 timestamp.
642 t, err = time.Parse(nearlyRFC1123, s)
647 // Untrash moves block from trash back into store
648 func (v *S3Volume) Untrash(loc string) error {
649 err := v.safeCopy(loc, "trash/"+loc)
653 err = v.bucket.PutReader("recent/"+loc, nil, 0, "application/octet-stream", s3ACL, s3.Options{})
654 return v.translateError(err)
657 // Status returns a *VolumeStatus representing the current in-use
658 // storage capacity and a fake available capacity that doesn't make
659 // the volume seem full or nearly-full.
660 func (v *S3Volume) Status() *VolumeStatus {
661 return &VolumeStatus{
663 BytesFree: BlockSize * 1000,
668 // InternalStats returns bucket I/O and API call counters.
669 func (v *S3Volume) InternalStats() interface{} {
670 return &v.bucket.stats
673 // String implements fmt.Stringer.
674 func (v *S3Volume) String() string {
675 return fmt.Sprintf("s3-bucket:%+q", v.Bucket)
678 // Writable returns false if all future Put, Mtime, and Delete calls
679 // are expected to fail.
680 func (v *S3Volume) Writable() bool {
684 // Replication returns the storage redundancy of the underlying
685 // device. Configured via command line flag.
686 func (v *S3Volume) Replication() int {
687 return v.S3Replication
690 var s3KeepBlockRegexp = regexp.MustCompile(`^[0-9a-f]{32}$`)
692 func (v *S3Volume) isKeepBlock(s string) bool {
693 return s3KeepBlockRegexp.MatchString(s)
696 // fixRace(X) is called when "recent/X" exists but "X" doesn't
697 // exist. If the timestamps on "recent/"+loc and "trash/"+loc indicate
698 // there was a race between Put and Trash, fixRace recovers from the
699 // race by Untrashing the block.
700 func (v *S3Volume) fixRace(loc string) bool {
701 trash, err := v.bucket.Head("trash/"+loc, nil)
703 if !os.IsNotExist(v.translateError(err)) {
704 log.Printf("error: fixRace: HEAD %q: %s", "trash/"+loc, err)
708 trashTime, err := v.lastModified(trash)
710 log.Printf("error: fixRace: parse %q: %s", trash.Header.Get("Last-Modified"), err)
714 recent, err := v.bucket.Head("recent/"+loc, nil)
716 log.Printf("error: fixRace: HEAD %q: %s", "recent/"+loc, err)
719 recentTime, err := v.lastModified(recent)
721 log.Printf("error: fixRace: parse %q: %s", recent.Header.Get("Last-Modified"), err)
725 ageWhenTrashed := trashTime.Sub(recentTime)
726 if ageWhenTrashed >= theConfig.BlobSignatureTTL.Duration() {
727 // No evidence of a race: block hasn't been written
728 // since it became eligible for Trash. No fix needed.
732 log.Printf("notice: fixRace: %q: trashed at %s but touched at %s (age when trashed = %s < %s)", loc, trashTime, recentTime, ageWhenTrashed, theConfig.BlobSignatureTTL)
733 log.Printf("notice: fixRace: copying %q to %q to recover from race between Put/Touch and Trash", "recent/"+loc, loc)
734 err = v.safeCopy(loc, "trash/"+loc)
736 log.Printf("error: fixRace: %s", err)
742 func (v *S3Volume) translateError(err error) error {
743 switch err := err.(type) {
745 if (err.StatusCode == http.StatusNotFound && err.Code == "NoSuchKey") ||
746 strings.Contains(err.Error(), "Not Found") {
747 return os.ErrNotExist
749 // Other 404 errors like NoSuchVersion and
750 // NoSuchBucket are different problems which should
751 // get called out downstream, so we don't convert them
752 // to os.ErrNotExist.
757 // EmptyTrash looks for trashed blocks that exceeded TrashLifetime
758 // and deletes them from the volume.
759 func (v *S3Volume) EmptyTrash() {
760 var bytesInTrash, blocksInTrash, bytesDeleted, blocksDeleted int64
762 // Use a merge sort to find matching sets of trash/X and recent/X.
764 Bucket: v.bucket.Bucket,
766 PageSize: v.IndexPageSize,
768 // Define "ready to delete" as "...when EmptyTrash started".
770 for trash := trashL.First(); trash != nil; trash = trashL.Next() {
772 if !v.isKeepBlock(loc) {
775 bytesInTrash += trash.Size
778 trashT, err := time.Parse(time.RFC3339, trash.LastModified)
780 log.Printf("warning: %s: EmptyTrash: %q: parse %q: %s", v, trash.Key, trash.LastModified, err)
783 recent, err := v.bucket.Head("recent/"+loc, nil)
784 if err != nil && os.IsNotExist(v.translateError(err)) {
785 log.Printf("warning: %s: EmptyTrash: found trash marker %q but no %q (%s); calling Untrash", v, trash.Key, "recent/"+loc, err)
788 log.Printf("error: %s: EmptyTrash: Untrash(%q): %s", v, loc, err)
791 } else if err != nil {
792 log.Printf("warning: %s: EmptyTrash: HEAD %q: %s", v, "recent/"+loc, err)
795 recentT, err := v.lastModified(recent)
797 log.Printf("warning: %s: EmptyTrash: %q: parse %q: %s", v, "recent/"+loc, recent.Header.Get("Last-Modified"), err)
800 if trashT.Sub(recentT) < theConfig.BlobSignatureTTL.Duration() {
801 if age := startT.Sub(recentT); age >= theConfig.BlobSignatureTTL.Duration()-time.Duration(v.RaceWindow) {
802 // recent/loc is too old to protect
803 // loc from being Trashed again during
804 // the raceWindow that starts if we
805 // delete trash/X now.
807 // Note this means (TrashCheckInterval
808 // < BlobSignatureTTL - raceWindow) is
809 // necessary to avoid starvation.
810 log.Printf("notice: %s: EmptyTrash: detected old race for %q, calling fixRace + Touch", v, loc)
815 _, err := v.bucket.Head(loc, nil)
816 if os.IsNotExist(err) {
817 log.Printf("notice: %s: EmptyTrash: detected recent race for %q, calling fixRace", v, loc)
820 } else if err != nil {
821 log.Printf("warning: %s: EmptyTrash: HEAD %q: %s", v, loc, err)
825 if startT.Sub(trashT) < theConfig.TrashLifetime.Duration() {
828 err = v.bucket.Del(trash.Key)
830 log.Printf("warning: %s: EmptyTrash: deleting %q: %s", v, trash.Key, err)
833 bytesDeleted += trash.Size
836 _, err = v.bucket.Head(loc, nil)
837 if os.IsNotExist(err) {
838 err = v.bucket.Del("recent/" + loc)
840 log.Printf("warning: %s: EmptyTrash: deleting %q: %s", v, "recent/"+loc, err)
842 } else if err != nil {
843 log.Printf("warning: %s: EmptyTrash: HEAD %q: %s", v, "recent/"+loc, err)
846 if err := trashL.Error(); err != nil {
847 log.Printf("error: %s: EmptyTrash: lister: %s", v, err)
849 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)
852 type s3Lister struct {
861 // First fetches the first page and returns the first item. It returns
862 // nil if the response is the empty set or an error occurs.
863 func (lister *s3Lister) First() *s3.Key {
868 // Next returns the next item, fetching the next page if necessary. It
869 // returns nil if the last available item has already been fetched, or
871 func (lister *s3Lister) Next() *s3.Key {
872 if len(lister.buf) == 0 && lister.nextMarker != "" {
878 // Return the most recent error encountered by First or Next.
879 func (lister *s3Lister) Error() error {
883 func (lister *s3Lister) getPage() {
884 resp, err := lister.Bucket.List(lister.Prefix, "", lister.nextMarker, lister.PageSize)
885 lister.nextMarker = ""
890 if resp.IsTruncated {
891 lister.nextMarker = resp.NextMarker
893 lister.buf = make([]s3.Key, 0, len(resp.Contents))
894 for _, key := range resp.Contents {
895 if !strings.HasPrefix(key.Key, lister.Prefix) {
896 log.Printf("warning: s3Lister: S3 Bucket.List(prefix=%q) returned key %q", lister.Prefix, key.Key)
899 lister.buf = append(lister.buf, key)
903 func (lister *s3Lister) pop() (k *s3.Key) {
904 if len(lister.buf) > 0 {
906 lister.buf = lister.buf[1:]
911 // s3bucket wraps s3.bucket and counts I/O and API usage stats.
912 type s3bucket struct {
917 func (b *s3bucket) GetReader(path string) (io.ReadCloser, error) {
918 rdr, err := b.Bucket.GetReader(path)
919 b.stats.Tick(&b.stats.Ops, &b.stats.GetOps)
921 return NewCountingReader(rdr, b.stats.TickInBytes), err
924 func (b *s3bucket) Head(path string, headers map[string][]string) (*http.Response, error) {
925 resp, err := b.Bucket.Head(path, headers)
926 b.stats.Tick(&b.stats.Ops, &b.stats.HeadOps)
931 func (b *s3bucket) PutReader(path string, r io.Reader, length int64, contType string, perm s3.ACL, options s3.Options) error {
933 // goamz will only send Content-Length: 0 when reader
934 // is nil due to net.http.Request.ContentLength
935 // behavior. Otherwise, Content-Length header is
936 // omitted which will cause some S3 services
937 // (including AWS and Ceph RadosGW) to fail to create
941 r = NewCountingReader(r, b.stats.TickOutBytes)
943 err := b.Bucket.PutReader(path, r, length, contType, perm, options)
944 b.stats.Tick(&b.stats.Ops, &b.stats.PutOps)
949 func (b *s3bucket) Del(path string) error {
950 err := b.Bucket.Del(path)
951 b.stats.Tick(&b.stats.Ops, &b.stats.DelOps)
956 type s3bucketStats struct {
966 func (s *s3bucketStats) TickErr(err error) {
970 errType := fmt.Sprintf("%T", err)
971 if err, ok := err.(*s3.Error); ok {
972 errType = errType + fmt.Sprintf(" %d %s", err.StatusCode, err.Code)
974 s.statsTicker.TickErr(err, errType)