X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/aa3efa4aa2749e9c20f6b889ce5968b84db283ba..7ebe828a435dcaa1b5668b72adbaad495059f211:/services/keepstore/unix_volume.go

diff --git a/services/keepstore/unix_volume.go b/services/keepstore/unix_volume.go
index 46f4db4095..dd62cf1319 100644
--- a/services/keepstore/unix_volume.go
+++ b/services/keepstore/unix_volume.go
@@ -321,7 +321,12 @@ func (v *UnixVolume) Status() *VolumeStatus {
 		v.logger.WithError(err).Error("stat failed")
 		return nil
 	}
-	devnum := fi.Sys().(*syscall.Stat_t).Dev
+	// uint64() cast here supports GOOS=darwin where Dev is
+	// int32. If the device number is negative, the unsigned
+	// devnum won't be the real device number any more, but that's
+	// fine -- all we care about is getting the same number each
+	// time.
+	devnum := uint64(fi.Sys().(*syscall.Stat_t).Dev)
 
 	var fs syscall.Statfs_t
 	if err := syscall.Statfs(v.Root, &fs); err != nil {
@@ -379,23 +384,25 @@ func (v *UnixVolume) IndexTo(prefix string, w io.Writer) error {
 			continue
 		}
 		blockdirpath := filepath.Join(v.Root, subdir)
-		blockdir, err := v.os.Open(blockdirpath)
-		if err != nil {
-			v.logger.WithError(err).Errorf("error reading %q", blockdirpath)
-			return fmt.Errorf("error reading %q: %s", blockdirpath, err)
-		}
-		v.os.stats.TickOps("readdir")
-		v.os.stats.Tick(&v.os.stats.ReaddirOps)
-		// ReadDir() (compared to Readdir(), which returns
-		// FileInfo structs) helps complete the sequence of
-		// readdirent calls as quickly as possible, reducing
-		// the likelihood of NFS EBADCOOKIE (523) errors.
-		dirents, err := blockdir.ReadDir(-1)
-		blockdir.Close()
-		if err != nil {
-			v.logger.WithError(err).Errorf("error reading %q", blockdirpath)
-			return fmt.Errorf("error reading %q: %s", blockdirpath, err)
+
+		var dirents []os.DirEntry
+		for attempt := 0; ; attempt++ {
+			v.os.stats.TickOps("readdir")
+			v.os.stats.Tick(&v.os.stats.ReaddirOps)
+			dirents, err = os.ReadDir(blockdirpath)
+			if err == nil {
+				break
+			} else if attempt < 5 && strings.Contains(err.Error(), "errno 523") {
+				// EBADCOOKIE (NFS stopped accepting
+				// our readdirent cookie) -- retry a
+				// few times before giving up
+				v.logger.WithError(err).Printf("retry after error reading %s", blockdirpath)
+				continue
+			} else {
+				return err
+			}
 		}
+
 		for _, dirent := range dirents {
 			fileInfo, err := dirent.Info()
 			if os.IsNotExist(err) {