1 // Copyright (C) The Arvados Authors. All rights reserved.
3 // SPDX-License-Identifier: AGPL-3.0
27 "git.arvados.org/arvados.git/sdk/go/arvados"
28 "github.com/AdRoll/goamz/aws"
29 "github.com/AdRoll/goamz/s3"
30 "github.com/prometheus/client_golang/prometheus"
31 "github.com/sirupsen/logrus"
35 driver["S3"] = newS3Volume
38 func newS3Volume(cluster *arvados.Cluster, volume arvados.Volume, logger logrus.FieldLogger, metrics *volumeMetricsVecs) (Volume, error) {
39 v := &S3Volume{cluster: cluster, volume: volume, metrics: metrics}
40 err := json.Unmarshal(volume.DriverParameters, &v)
44 v.logger = logger.WithField("Volume", v.String())
48 func (v *S3Volume) check() error {
50 return errors.New("DriverParameters: Bucket must be provided")
52 if v.IndexPageSize == 0 {
53 v.IndexPageSize = 1000
56 return errors.New("DriverParameters: RaceWindow must not be negative")
60 v.region, ok = aws.Regions[v.Region]
63 return fmt.Errorf("unrecognized region %+q; try specifying endpoint instead", v.Region)
66 return fmt.Errorf("refusing to use AWS region name %+q with endpoint %+q; "+
67 "specify empty endpoint or use a different region name", v.Region, v.Endpoint)
69 v.region = aws.Region{
71 S3Endpoint: v.Endpoint,
72 S3LocationConstraint: v.LocationConstraint,
76 // Zero timeouts mean "wait forever", which is a bad
77 // default. Default to long timeouts instead.
78 if v.ConnectTimeout == 0 {
79 v.ConnectTimeout = s3DefaultConnectTimeout
81 if v.ReadTimeout == 0 {
82 v.ReadTimeout = s3DefaultReadTimeout
91 // Set up prometheus metrics
92 lbls := prometheus.Labels{"device_id": v.GetDeviceID()}
93 v.bucket.stats.opsCounters, v.bucket.stats.errCounters, v.bucket.stats.ioBytes = v.metrics.getCounterVecsFor(lbls)
95 err := v.bootstrapIAMCredentials()
97 return fmt.Errorf("error getting IAM credentials: %s", err)
104 s3DefaultReadTimeout = arvados.Duration(10 * time.Minute)
105 s3DefaultConnectTimeout = arvados.Duration(time.Minute)
109 // ErrS3TrashDisabled is returned by Trash if that operation
110 // is impossible with the current config.
111 ErrS3TrashDisabled = fmt.Errorf("trash function is disabled because Collections.BlobTrashLifetime=0 and DriverParameters.UnsafeDelete=false")
119 maxClockSkew = 600 * time.Second
120 nearlyRFC1123 = "Mon, 2 Jan 2006 15:04:05 GMT"
123 func s3regions() (okList []string) {
124 for r := range aws.Regions {
125 okList = append(okList, r)
130 // S3Volume implements Volume using an S3 bucket.
131 type S3Volume struct {
134 AuthToken string // populated automatically when IAMRole is used
135 AuthExpiration time.Time // populated automatically when IAMRole is used
140 LocationConstraint bool
142 ConnectTimeout arvados.Duration
143 ReadTimeout arvados.Duration
144 RaceWindow arvados.Duration
147 cluster *arvados.Cluster
148 volume arvados.Volume
149 logger logrus.FieldLogger
150 metrics *volumeMetricsVecs
156 // GetDeviceID returns a globally unique ID for the storage bucket.
157 func (v *S3Volume) GetDeviceID() string {
158 return "s3://" + v.Endpoint + "/" + v.Bucket
161 func (v *S3Volume) bootstrapIAMCredentials() error {
162 if v.AccessKey != "" || v.SecretKey != "" {
164 return errors.New("invalid DriverParameters: AccessKey and SecretKey must be blank if IAMRole is specified")
168 ttl, err := v.updateIAMCredentials()
175 ttl, err = v.updateIAMCredentials()
177 v.logger.WithError(err).Warnf("failed to update credentials for IAM role %q", v.IAMRole)
179 } else if ttl < time.Second {
180 v.logger.WithField("TTL", ttl).Warnf("received stale credentials for IAM role %q", v.IAMRole)
188 func (v *S3Volume) newS3Client() *s3.S3 {
189 auth := aws.NewAuth(v.AccessKey, v.SecretKey, v.AuthToken, v.AuthExpiration)
190 client := s3.New(*auth, v.region)
191 if v.region.EC2Endpoint.Signer == aws.V4Signature {
192 // Currently affects only eu-central-1
193 client.Signature = aws.V4Signature
195 client.ConnectTimeout = time.Duration(v.ConnectTimeout)
196 client.ReadTimeout = time.Duration(v.ReadTimeout)
200 // returned by AWS metadata endpoint .../security-credentials/${rolename}
201 type iamCredentials struct {
203 LastUpdated time.Time
206 SecretAccessKey string
211 // Returns TTL of updated credentials, i.e., time to sleep until next
213 func (v *S3Volume) updateIAMCredentials() (time.Duration, error) {
214 ctx, cancel := context.WithDeadline(context.Background(), time.Now().Add(time.Minute))
217 metadataBaseURL := "http://169.254.169.254/latest/meta-data/iam/security-credentials/"
220 if strings.Contains(v.IAMRole, "://") {
221 // Configuration provides complete URL (used by tests)
223 } else if v.IAMRole != "" {
224 // Configuration provides IAM role name and we use the
225 // AWS metadata endpoint
226 url = metadataBaseURL + v.IAMRole
228 url = metadataBaseURL
229 v.logger.WithField("URL", url).Debug("looking up IAM role name")
230 req, err := http.NewRequest("GET", url, nil)
232 return 0, fmt.Errorf("error setting up request %s: %s", url, err)
234 resp, err := http.DefaultClient.Do(req.WithContext(ctx))
236 return 0, fmt.Errorf("error getting %s: %s", url, err)
238 defer resp.Body.Close()
239 if resp.StatusCode == http.StatusNotFound {
240 return 0, fmt.Errorf("this instance does not have an IAM role assigned -- either assign a role, or configure AccessKey and SecretKey explicitly in DriverParameters (error getting %s: HTTP status %s)", url, resp.Status)
241 } else if resp.StatusCode != http.StatusOK {
242 return 0, fmt.Errorf("error getting %s: HTTP status %s", url, resp.Status)
244 body := bufio.NewReader(resp.Body)
246 _, err = fmt.Fscanf(body, "%s\n", &role)
248 return 0, fmt.Errorf("error reading response from %s: %s", url, err)
250 if n, _ := body.Read(make([]byte, 64)); n > 0 {
251 v.logger.Warnf("ignoring additional data returned by metadata endpoint %s after the single role name that we expected", url)
253 v.logger.WithField("Role", role).Debug("looked up IAM role name")
257 v.logger.WithField("URL", url).Debug("getting credentials")
258 req, err := http.NewRequest("GET", url, nil)
260 return 0, fmt.Errorf("error setting up request %s: %s", url, err)
262 resp, err := http.DefaultClient.Do(req.WithContext(ctx))
264 return 0, fmt.Errorf("error getting %s: %s", url, err)
266 defer resp.Body.Close()
267 if resp.StatusCode != http.StatusOK {
268 return 0, fmt.Errorf("error getting %s: HTTP status %s", url, resp.Status)
270 var cred iamCredentials
271 err = json.NewDecoder(resp.Body).Decode(&cred)
273 return 0, fmt.Errorf("error decoding credentials from %s: %s", url, err)
275 v.AccessKey, v.SecretKey, v.AuthToken, v.AuthExpiration = cred.AccessKeyID, cred.SecretAccessKey, cred.Token, cred.Expiration
276 v.bucket.SetBucket(&s3.Bucket{
280 // TTL is time from now to expiration, minus 5m. "We make new
281 // credentials available at least five minutes before the
282 // expiration of the old credentials." --
283 // https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/iam-roles-for-amazon-ec2.html#instance-metadata-security-credentials
284 // (If that's not true, the returned ttl might be zero or
285 // negative, which the caller can handle.)
286 ttl := cred.Expiration.Sub(time.Now()) - 5*time.Minute
287 v.logger.WithFields(logrus.Fields{
288 "AccessKeyID": cred.AccessKeyID,
289 "LastUpdated": cred.LastUpdated,
290 "Expiration": cred.Expiration,
291 "TTL": arvados.Duration(ttl),
292 }).Debug("updated credentials")
296 func (v *S3Volume) getReaderWithContext(ctx context.Context, loc string) (rdr io.ReadCloser, err error) {
297 ready := make(chan bool)
299 rdr, err = v.getReader(loc)
306 v.logger.Debugf("s3: abandoning getReader(): %s", ctx.Err())
313 return nil, ctx.Err()
317 // getReader wraps (Bucket)GetReader.
319 // In situations where (Bucket)GetReader would fail because the block
320 // disappeared in a Trash race, getReader calls fixRace to recover the
321 // data, and tries again.
322 func (v *S3Volume) getReader(loc string) (rdr io.ReadCloser, err error) {
323 rdr, err = v.bucket.GetReader(loc)
324 err = v.translateError(err)
325 if err == nil || !os.IsNotExist(err) {
329 _, err = v.bucket.Head("recent/"+loc, nil)
330 err = v.translateError(err)
332 // If we can't read recent/X, there's no point in
333 // trying fixRace. Give up.
341 rdr, err = v.bucket.GetReader(loc)
343 v.logger.Warnf("reading %s after successful fixRace: %s", loc, err)
344 err = v.translateError(err)
349 // Get a block: copy the block data into buf, and return the number of
351 func (v *S3Volume) Get(ctx context.Context, loc string, buf []byte) (int, error) {
352 rdr, err := v.getReaderWithContext(ctx, loc)
358 ready := make(chan bool)
363 n, err = io.ReadFull(rdr, buf)
366 case nil, io.EOF, io.ErrUnexpectedEOF:
369 err = v.translateError(err)
374 v.logger.Debugf("s3: interrupting ReadFull() with Close() because %s", ctx.Err())
376 // Must wait for ReadFull to return, to ensure it
377 // doesn't write to buf after we return.
378 v.logger.Debug("s3: waiting for ReadFull() to fail")
386 // Compare the given data with the stored data.
387 func (v *S3Volume) Compare(ctx context.Context, loc string, expect []byte) error {
388 errChan := make(chan error, 1)
390 _, err := v.bucket.Head("recent/"+loc, nil)
397 case err = <-errChan:
400 // Checking for "loc" itself here would interfere with
401 // future GET requests.
403 // On AWS, if X doesn't exist, a HEAD or GET request
404 // for X causes X's non-existence to be cached. Thus,
405 // if we test for X, then create X and return a
406 // signature to our client, the client might still get
407 // 404 from all keepstores when trying to read it.
409 // To avoid this, we avoid doing HEAD X or GET X until
410 // we know X has been written.
412 // Note that X might exist even though recent/X
413 // doesn't: for example, the response to HEAD recent/X
414 // might itself come from a stale cache. In such
415 // cases, we will return a false negative and
416 // PutHandler might needlessly create another replica
417 // on a different volume. That's not ideal, but it's
418 // better than passing the eventually-consistent
419 // problem on to our clients.
420 return v.translateError(err)
422 rdr, err := v.getReaderWithContext(ctx, loc)
427 return v.translateError(compareReaderWithBuf(ctx, rdr, expect, loc[:32]))
430 // Put writes a block.
431 func (v *S3Volume) Put(ctx context.Context, loc string, block []byte) error {
432 if v.volume.ReadOnly {
433 return MethodDisabledError
438 md5, err := hex.DecodeString(loc)
442 opts.ContentMD5 = base64.StdEncoding.EncodeToString(md5)
443 // In AWS regions that use V4 signatures, we need to
444 // provide ContentSHA256 up front. Otherwise, the S3
445 // library reads the request body (from our buffer)
446 // into another new buffer in order to compute the
447 // SHA256 before sending the request -- which would
448 // mean consuming 128 MiB of memory for the duration
449 // of a 64 MiB write.
450 opts.ContentSHA256 = fmt.Sprintf("%x", sha256.Sum256(block))
453 // Send the block data through a pipe, so that (if we need to)
454 // we can close the pipe early and abandon our PutReader()
455 // goroutine, without worrying about PutReader() accessing our
456 // block buffer after we release it.
457 bufr, bufw := io.Pipe()
459 io.Copy(bufw, bytes.NewReader(block))
464 ready := make(chan bool)
467 if ctx.Err() != nil {
468 v.logger.Debugf("abandoned PutReader goroutine finished with err: %s", err)
472 err = v.bucket.PutReader(loc, bufr, int64(size), "application/octet-stream", s3ACL, opts)
476 err = v.bucket.PutReader("recent/"+loc, nil, 0, "application/octet-stream", s3ACL, s3.Options{})
480 v.logger.Debugf("taking PutReader's input away: %s", ctx.Err())
481 // Our pipe might be stuck in Write(), waiting for
482 // PutReader() to read. If so, un-stick it. This means
483 // PutReader will get corrupt data, but that's OK: the
484 // size and MD5 won't match, so the write will fail.
485 go io.Copy(ioutil.Discard, bufr)
486 // CloseWithError() will return once pending I/O is done.
487 bufw.CloseWithError(ctx.Err())
488 v.logger.Debugf("abandoning PutReader goroutine")
491 // Unblock pipe in case PutReader did not consume it.
492 io.Copy(ioutil.Discard, bufr)
493 return v.translateError(err)
497 // Touch sets the timestamp for the given locator to the current time.
498 func (v *S3Volume) Touch(loc string) error {
499 if v.volume.ReadOnly {
500 return MethodDisabledError
502 _, err := v.bucket.Head(loc, nil)
503 err = v.translateError(err)
504 if os.IsNotExist(err) && v.fixRace(loc) {
505 // The data object got trashed in a race, but fixRace
507 } else if err != nil {
510 err = v.bucket.PutReader("recent/"+loc, nil, 0, "application/octet-stream", s3ACL, s3.Options{})
511 return v.translateError(err)
514 // Mtime returns the stored timestamp for the given locator.
515 func (v *S3Volume) Mtime(loc string) (time.Time, error) {
516 _, err := v.bucket.Head(loc, nil)
518 return zeroTime, v.translateError(err)
520 resp, err := v.bucket.Head("recent/"+loc, nil)
521 err = v.translateError(err)
522 if os.IsNotExist(err) {
523 // The data object X exists, but recent/X is missing.
524 err = v.bucket.PutReader("recent/"+loc, nil, 0, "application/octet-stream", s3ACL, s3.Options{})
526 v.logger.WithError(err).Errorf("error creating %q", "recent/"+loc)
527 return zeroTime, v.translateError(err)
529 v.logger.Infof("created %q to migrate existing block to new storage scheme", "recent/"+loc)
530 resp, err = v.bucket.Head("recent/"+loc, nil)
532 v.logger.WithError(err).Errorf("HEAD failed after creating %q", "recent/"+loc)
533 return zeroTime, v.translateError(err)
535 } else if err != nil {
536 // HEAD recent/X failed for some other reason.
539 return v.lastModified(resp)
542 // IndexTo writes a complete list of locators with the given prefix
543 // for which Get() can retrieve data.
544 func (v *S3Volume) IndexTo(prefix string, writer io.Writer) error {
545 // Use a merge sort to find matching sets of X and recent/X.
548 Bucket: v.bucket.Bucket(),
550 PageSize: v.IndexPageSize,
551 Stats: &v.bucket.stats,
555 Bucket: v.bucket.Bucket(),
556 Prefix: "recent/" + prefix,
557 PageSize: v.IndexPageSize,
558 Stats: &v.bucket.stats,
560 for data, recent := dataL.First(), recentL.First(); data != nil && dataL.Error() == nil; data = dataL.Next() {
562 // Conveniently, "recent/*" and "trash/*" are
563 // lexically greater than all hex-encoded data
564 // hashes, so stopping here avoids iterating
565 // over all of them needlessly with dataL.
568 if !v.isKeepBlock(data.Key) {
572 // stamp is the list entry we should use to report the
573 // last-modified time for this data block: it will be
574 // the recent/X entry if one exists, otherwise the
575 // entry for the data block itself.
578 // Advance to the corresponding recent/X marker, if any
579 for recent != nil && recentL.Error() == nil {
580 if cmp := strings.Compare(recent.Key[7:], data.Key); cmp < 0 {
581 recent = recentL.Next()
585 recent = recentL.Next()
588 // recent/X marker is missing: we'll
589 // use the timestamp on the data
594 if err := recentL.Error(); err != nil {
597 t, err := time.Parse(time.RFC3339, stamp.LastModified)
601 fmt.Fprintf(writer, "%s+%d %d\n", data.Key, data.Size, t.UnixNano())
606 // Trash a Keep block.
607 func (v *S3Volume) Trash(loc string) error {
608 if v.volume.ReadOnly {
609 return MethodDisabledError
611 if t, err := v.Mtime(loc); err != nil {
613 } else if time.Since(t) < v.cluster.Collections.BlobSigningTTL.Duration() {
616 if v.cluster.Collections.BlobTrashLifetime == 0 {
618 return ErrS3TrashDisabled
620 return v.translateError(v.bucket.Del(loc))
622 err := v.checkRaceWindow(loc)
626 err = v.safeCopy("trash/"+loc, loc)
630 return v.translateError(v.bucket.Del(loc))
633 // checkRaceWindow returns a non-nil error if trash/loc is, or might
634 // be, in the race window (i.e., it's not safe to trash loc).
635 func (v *S3Volume) checkRaceWindow(loc string) error {
636 resp, err := v.bucket.Head("trash/"+loc, nil)
637 err = v.translateError(err)
638 if os.IsNotExist(err) {
639 // OK, trash/X doesn't exist so we're not in the race
642 } else if err != nil {
643 // Error looking up trash/X. We don't know whether
644 // we're in the race window
647 t, err := v.lastModified(resp)
649 // Can't parse timestamp
652 safeWindow := t.Add(v.cluster.Collections.BlobTrashLifetime.Duration()).Sub(time.Now().Add(time.Duration(v.RaceWindow)))
654 // We can't count on "touch trash/X" to prolong
655 // trash/X's lifetime. The new timestamp might not
656 // become visible until now+raceWindow, and EmptyTrash
657 // is allowed to delete trash/X before then.
658 return fmt.Errorf("same block is already in trash, and safe window ended %s ago", -safeWindow)
660 // trash/X exists, but it won't be eligible for deletion until
661 // after now+raceWindow, so it's safe to overwrite it.
665 // safeCopy calls PutCopy, and checks the response to make sure the
666 // copy succeeded and updated the timestamp on the destination object
667 // (PutCopy returns 200 OK if the request was received, even if the
669 func (v *S3Volume) safeCopy(dst, src string) error {
670 resp, err := v.bucket.Bucket().PutCopy(dst, s3ACL, s3.CopyOptions{
671 ContentType: "application/octet-stream",
672 MetadataDirective: "REPLACE",
673 }, v.bucket.Bucket().Name+"/"+src)
674 err = v.translateError(err)
675 if os.IsNotExist(err) {
677 } else if err != nil {
678 return fmt.Errorf("PutCopy(%q ← %q): %s", dst, v.bucket.Bucket().Name+"/"+src, err)
680 if t, err := time.Parse(time.RFC3339Nano, resp.LastModified); err != nil {
681 return fmt.Errorf("PutCopy succeeded but did not return a timestamp: %q: %s", resp.LastModified, err)
682 } else if time.Now().Sub(t) > maxClockSkew {
683 return fmt.Errorf("PutCopy succeeded but returned an old timestamp: %q: %s", resp.LastModified, t)
688 // Get the LastModified header from resp, and parse it as RFC1123 or
689 // -- if it isn't valid RFC1123 -- as Amazon's variant of RFC1123.
690 func (v *S3Volume) lastModified(resp *http.Response) (t time.Time, err error) {
691 s := resp.Header.Get("Last-Modified")
692 t, err = time.Parse(time.RFC1123, s)
693 if err != nil && s != "" {
694 // AWS example is "Sun, 1 Jan 2006 12:00:00 GMT",
695 // which isn't quite "Sun, 01 Jan 2006 12:00:00 GMT"
696 // as required by HTTP spec. If it's not a valid HTTP
697 // header value, it's probably AWS (or s3test) giving
698 // us a nearly-RFC1123 timestamp.
699 t, err = time.Parse(nearlyRFC1123, s)
704 // Untrash moves block from trash back into store
705 func (v *S3Volume) Untrash(loc string) error {
706 err := v.safeCopy(loc, "trash/"+loc)
710 err = v.bucket.PutReader("recent/"+loc, nil, 0, "application/octet-stream", s3ACL, s3.Options{})
711 return v.translateError(err)
714 // Status returns a *VolumeStatus representing the current in-use
715 // storage capacity and a fake available capacity that doesn't make
716 // the volume seem full or nearly-full.
717 func (v *S3Volume) Status() *VolumeStatus {
718 return &VolumeStatus{
720 BytesFree: BlockSize * 1000,
725 // InternalStats returns bucket I/O and API call counters.
726 func (v *S3Volume) InternalStats() interface{} {
727 return &v.bucket.stats
730 // String implements fmt.Stringer.
731 func (v *S3Volume) String() string {
732 return fmt.Sprintf("s3-bucket:%+q", v.Bucket)
735 var s3KeepBlockRegexp = regexp.MustCompile(`^[0-9a-f]{32}$`)
737 func (v *S3Volume) isKeepBlock(s string) bool {
738 return s3KeepBlockRegexp.MatchString(s)
741 // fixRace(X) is called when "recent/X" exists but "X" doesn't
742 // exist. If the timestamps on "recent/"+loc and "trash/"+loc indicate
743 // there was a race between Put and Trash, fixRace recovers from the
744 // race by Untrashing the block.
745 func (v *S3Volume) fixRace(loc string) bool {
746 trash, err := v.bucket.Head("trash/"+loc, nil)
748 if !os.IsNotExist(v.translateError(err)) {
749 v.logger.WithError(err).Errorf("fixRace: HEAD %q failed", "trash/"+loc)
753 trashTime, err := v.lastModified(trash)
755 v.logger.WithError(err).Errorf("fixRace: error parsing time %q", trash.Header.Get("Last-Modified"))
759 recent, err := v.bucket.Head("recent/"+loc, nil)
761 v.logger.WithError(err).Errorf("fixRace: HEAD %q failed", "recent/"+loc)
764 recentTime, err := v.lastModified(recent)
766 v.logger.WithError(err).Errorf("fixRace: error parsing time %q", recent.Header.Get("Last-Modified"))
770 ageWhenTrashed := trashTime.Sub(recentTime)
771 if ageWhenTrashed >= v.cluster.Collections.BlobSigningTTL.Duration() {
772 // No evidence of a race: block hasn't been written
773 // since it became eligible for Trash. No fix needed.
777 v.logger.Infof("fixRace: %q: trashed at %s but touched at %s (age when trashed = %s < %s)", loc, trashTime, recentTime, ageWhenTrashed, v.cluster.Collections.BlobSigningTTL)
778 v.logger.Infof("fixRace: copying %q to %q to recover from race between Put/Touch and Trash", "recent/"+loc, loc)
779 err = v.safeCopy(loc, "trash/"+loc)
781 v.logger.WithError(err).Error("fixRace: copy failed")
787 func (v *S3Volume) translateError(err error) error {
788 switch err := err.(type) {
790 if (err.StatusCode == http.StatusNotFound && err.Code == "NoSuchKey") ||
791 strings.Contains(err.Error(), "Not Found") {
792 return os.ErrNotExist
794 // Other 404 errors like NoSuchVersion and
795 // NoSuchBucket are different problems which should
796 // get called out downstream, so we don't convert them
797 // to os.ErrNotExist.
802 // EmptyTrash looks for trashed blocks that exceeded BlobTrashLifetime
803 // and deletes them from the volume.
804 func (v *S3Volume) EmptyTrash() {
805 if v.cluster.Collections.BlobDeleteConcurrency < 1 {
809 var bytesInTrash, blocksInTrash, bytesDeleted, blocksDeleted int64
811 // Define "ready to delete" as "...when EmptyTrash started".
814 emptyOneKey := func(trash *s3.Key) {
816 if !v.isKeepBlock(loc) {
819 atomic.AddInt64(&bytesInTrash, trash.Size)
820 atomic.AddInt64(&blocksInTrash, 1)
822 trashT, err := time.Parse(time.RFC3339, trash.LastModified)
824 v.logger.Warnf("EmptyTrash: %q: parse %q: %s", trash.Key, trash.LastModified, err)
827 recent, err := v.bucket.Head("recent/"+loc, nil)
828 if err != nil && os.IsNotExist(v.translateError(err)) {
829 v.logger.Warnf("EmptyTrash: found trash marker %q but no %q (%s); calling Untrash", trash.Key, "recent/"+loc, err)
832 v.logger.WithError(err).Errorf("EmptyTrash: Untrash(%q) failed", loc)
835 } else if err != nil {
836 v.logger.WithError(err).Warnf("EmptyTrash: HEAD %q failed", "recent/"+loc)
839 recentT, err := v.lastModified(recent)
841 v.logger.WithError(err).Warnf("EmptyTrash: %q: error parsing %q", "recent/"+loc, recent.Header.Get("Last-Modified"))
844 if trashT.Sub(recentT) < v.cluster.Collections.BlobSigningTTL.Duration() {
845 if age := startT.Sub(recentT); age >= v.cluster.Collections.BlobSigningTTL.Duration()-time.Duration(v.RaceWindow) {
846 // recent/loc is too old to protect
847 // loc from being Trashed again during
848 // the raceWindow that starts if we
849 // delete trash/X now.
851 // Note this means (TrashSweepInterval
852 // < BlobSigningTTL - raceWindow) is
853 // necessary to avoid starvation.
854 v.logger.Infof("EmptyTrash: detected old race for %q, calling fixRace + Touch", loc)
859 _, err := v.bucket.Head(loc, nil)
860 if os.IsNotExist(err) {
861 v.logger.Infof("EmptyTrash: detected recent race for %q, calling fixRace", loc)
864 } else if err != nil {
865 v.logger.WithError(err).Warnf("EmptyTrash: HEAD %q failed", loc)
869 if startT.Sub(trashT) < v.cluster.Collections.BlobTrashLifetime.Duration() {
872 err = v.bucket.Del(trash.Key)
874 v.logger.WithError(err).Errorf("EmptyTrash: error deleting %q", trash.Key)
877 atomic.AddInt64(&bytesDeleted, trash.Size)
878 atomic.AddInt64(&blocksDeleted, 1)
880 _, err = v.bucket.Head(loc, nil)
882 v.logger.Warnf("EmptyTrash: HEAD %q succeeded immediately after deleting %q", loc, loc)
885 if !os.IsNotExist(v.translateError(err)) {
886 v.logger.WithError(err).Warnf("EmptyTrash: HEAD %q failed", loc)
889 err = v.bucket.Del("recent/" + loc)
891 v.logger.WithError(err).Warnf("EmptyTrash: error deleting %q", "recent/"+loc)
895 var wg sync.WaitGroup
896 todo := make(chan *s3.Key, v.cluster.Collections.BlobDeleteConcurrency)
897 for i := 0; i < v.cluster.Collections.BlobDeleteConcurrency; i++ {
901 for key := range todo {
909 Bucket: v.bucket.Bucket(),
911 PageSize: v.IndexPageSize,
912 Stats: &v.bucket.stats,
914 for trash := trashL.First(); trash != nil; trash = trashL.Next() {
920 if err := trashL.Error(); err != nil {
921 v.logger.WithError(err).Error("EmptyTrash: lister failed")
923 v.logger.Infof("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)
926 type s3Lister struct {
927 Logger logrus.FieldLogger
937 // First fetches the first page and returns the first item. It returns
938 // nil if the response is the empty set or an error occurs.
939 func (lister *s3Lister) First() *s3.Key {
944 // Next returns the next item, fetching the next page if necessary. It
945 // returns nil if the last available item has already been fetched, or
947 func (lister *s3Lister) Next() *s3.Key {
948 if len(lister.buf) == 0 && lister.nextMarker != "" {
954 // Return the most recent error encountered by First or Next.
955 func (lister *s3Lister) Error() error {
959 func (lister *s3Lister) getPage() {
960 lister.Stats.TickOps("list")
961 lister.Stats.Tick(&lister.Stats.Ops, &lister.Stats.ListOps)
962 resp, err := lister.Bucket.List(lister.Prefix, "", lister.nextMarker, lister.PageSize)
963 lister.nextMarker = ""
968 if resp.IsTruncated {
969 lister.nextMarker = resp.NextMarker
971 lister.buf = make([]s3.Key, 0, len(resp.Contents))
972 for _, key := range resp.Contents {
973 if !strings.HasPrefix(key.Key, lister.Prefix) {
974 lister.Logger.Warnf("s3Lister: S3 Bucket.List(prefix=%q) returned key %q", lister.Prefix, key.Key)
977 lister.buf = append(lister.buf, key)
981 func (lister *s3Lister) pop() (k *s3.Key) {
982 if len(lister.buf) > 0 {
984 lister.buf = lister.buf[1:]
989 // s3bucket wraps s3.bucket and counts I/O and API usage stats. The
990 // wrapped bucket can be replaced atomically with SetBucket in order
991 // to update credentials.
992 type s3bucket struct {
998 func (b *s3bucket) Bucket() *s3.Bucket {
1004 func (b *s3bucket) SetBucket(bucket *s3.Bucket) {
1010 func (b *s3bucket) GetReader(path string) (io.ReadCloser, error) {
1011 rdr, err := b.Bucket().GetReader(path)
1012 b.stats.TickOps("get")
1013 b.stats.Tick(&b.stats.Ops, &b.stats.GetOps)
1014 b.stats.TickErr(err)
1015 return NewCountingReader(rdr, b.stats.TickInBytes), err
1018 func (b *s3bucket) Head(path string, headers map[string][]string) (*http.Response, error) {
1019 resp, err := b.Bucket().Head(path, headers)
1020 b.stats.TickOps("head")
1021 b.stats.Tick(&b.stats.Ops, &b.stats.HeadOps)
1022 b.stats.TickErr(err)
1026 func (b *s3bucket) PutReader(path string, r io.Reader, length int64, contType string, perm s3.ACL, options s3.Options) error {
1028 // goamz will only send Content-Length: 0 when reader
1029 // is nil due to net.http.Request.ContentLength
1030 // behavior. Otherwise, Content-Length header is
1031 // omitted which will cause some S3 services
1032 // (including AWS and Ceph RadosGW) to fail to create
1036 r = NewCountingReader(r, b.stats.TickOutBytes)
1038 err := b.Bucket().PutReader(path, r, length, contType, perm, options)
1039 b.stats.TickOps("put")
1040 b.stats.Tick(&b.stats.Ops, &b.stats.PutOps)
1041 b.stats.TickErr(err)
1045 func (b *s3bucket) Del(path string) error {
1046 err := b.Bucket().Del(path)
1047 b.stats.TickOps("delete")
1048 b.stats.Tick(&b.stats.Ops, &b.stats.DelOps)
1049 b.stats.TickErr(err)
1053 type s3bucketStats struct {
1063 func (s *s3bucketStats) TickErr(err error) {
1067 errType := fmt.Sprintf("%T", err)
1068 if err, ok := err.(*s3.Error); ok {
1069 errType = errType + fmt.Sprintf(" %d %s", err.StatusCode, err.Code)
1071 s.statsTicker.TickErr(err, errType)