XL: Do not rely on getLoadBalancedQuorumDisks for NS consistency. (#2243)

The reason is any function relying on `getLoadBalancedQuorumDisks`
cannot possibly have an idempotent behavior.

The problem comes from given a set of N disks returning just a
shuffled N/2 disks.  In case of a scenario where we have N/2
number of failed disks, the returned value of `getLoadBalancedQuorumDisks`
is not equal to the same failed disks so essentially calls using such
disks might succeed or fail randomly at different intervals in time.

This proposal change is we move to `getLoadBalancedDisks()`
and use the shuffled N disks as a whole. Since most of the time we might
hit a good disk since we are not reducing our solution space. This
also provides consistent behavior for all the functions which rely
on shuffled disks.

Fixes #2242
master
Harshavardhana 8 years ago committed by Anand Babu (AB) Periasamy
parent 41f4f2806d
commit a0635dcdd9
  1. 72
      xl-v1-bucket.go
  2. 12
      xl-v1-common.go
  3. 2
      xl-v1-list-objects.go
  4. 2
      xl-v1-metadata.go
  5. 6
      xl-v1-multipart-common.go
  6. 10
      xl-v1-multipart.go

@ -30,6 +30,10 @@ func (xl xlObjects) MakeBucket(bucket string) error {
if !IsValidBucketName(bucket) {
return BucketNameInvalid{Bucket: bucket}
}
// Verify if bucket is found.
if xl.isBucketExist(bucket) {
return toObjectErr(errVolumeExists, bucket)
}
nsMutex.Lock(bucket, "")
defer nsMutex.Unlock(bucket, "")
@ -68,20 +72,36 @@ func (xl xlObjects) MakeBucket(bucket string) error {
}
// Verify we have any other errors which should undo make bucket.
for _, err := range dErrs {
// Bucket already exists, return BucketExists error.
if err == errVolumeExists {
return toObjectErr(errVolumeExists, bucket)
}
// Undo make bucket for any other errors.
if err != nil && err != errDiskNotFound {
xl.undoMakeBucket(bucket)
return toObjectErr(err, bucket)
}
if reducedErr := reduceErrs(dErrs, []error{
errDiskNotFound,
errFaultyDisk,
errDiskAccessDenied,
}); reducedErr != nil {
return toObjectErr(reducedErr, bucket)
}
return nil
}
func (xl xlObjects) undoDeleteBucket(bucket string) {
// Initialize sync waitgroup.
var wg = &sync.WaitGroup{}
// Undo previous make bucket entry on all underlying storage disks.
for index, disk := range xl.storageDisks {
if disk == nil {
continue
}
wg.Add(1)
// Delete a bucket inside a go-routine.
go func(index int, disk StorageAPI) {
defer wg.Done()
_ = disk.MakeVol(bucket)
}(index, disk)
}
// Wait for all make vol to finish.
wg.Wait()
}
// undo make bucket operation upon quorum failure.
func (xl xlObjects) undoMakeBucket(bucket string) {
// Initialize sync waitgroup.
@ -113,7 +133,7 @@ var bucketMetadataOpIgnoredErrs = []error{
// getBucketInfo - returns the BucketInfo from one of the load balanced disks.
func (xl xlObjects) getBucketInfo(bucketName string) (bucketInfo BucketInfo, err error) {
for _, disk := range xl.getLoadBalancedQuorumDisks() {
for _, disk := range xl.getLoadBalancedDisks() {
if disk == nil {
continue
}
@ -169,7 +189,7 @@ func (xl xlObjects) GetBucketInfo(bucket string) (BucketInfo, error) {
// listBuckets - returns list of all buckets from a disk picked at random.
func (xl xlObjects) listBuckets() (bucketsInfo []BucketInfo, err error) {
for _, disk := range xl.getLoadBalancedQuorumDisks() {
for _, disk := range xl.getLoadBalancedDisks() {
if disk == nil {
continue
}
@ -220,13 +240,15 @@ func (xl xlObjects) DeleteBucket(bucket string) error {
if !IsValidBucketName(bucket) {
return BucketNameInvalid{Bucket: bucket}
}
// Verify if bucket is found.
if !xl.isBucketExist(bucket) {
return BucketNotFound{Bucket: bucket}
}
nsMutex.Lock(bucket, "")
defer nsMutex.Unlock(bucket, "")
// Collect if all disks report volume not found.
var volumeNotFoundErrCnt int
var wg = &sync.WaitGroup{}
var dErrs = make([]error, len(xl.storageDisks))
@ -257,21 +279,17 @@ func (xl xlObjects) DeleteBucket(bucket string) error {
// Wait for all the delete vols to finish.
wg.Wait()
// Count the errors for known errors, return quickly if we found an unknown error.
for _, err := range dErrs {
if err != nil {
if isErrIgnored(err, objMetadataOpIgnoredErrs) {
volumeNotFoundErrCnt++
continue
}
return toObjectErr(err, bucket)
}
if !isDiskQuorum(dErrs, xl.writeQuorum) {
xl.undoDeleteBucket(bucket)
return toObjectErr(errXLWriteQuorum, bucket)
}
// Return errVolumeNotFound if all disks report volume not found.
if volumeNotFoundErrCnt == len(xl.storageDisks) {
return toObjectErr(errVolumeNotFound, bucket)
if reducedErr := reduceErrs(dErrs, []error{
errFaultyDisk,
errDiskNotFound,
errDiskAccessDenied,
}); reducedErr != nil {
return toObjectErr(reducedErr, bucket)
}
return nil
}

@ -21,15 +21,7 @@ import (
"time"
)
// getLoadBalancedQuorumDisks - fetches load balanced sufficiently
// randomized quorum disk slice.
func (xl xlObjects) getLoadBalancedQuorumDisks() (disks []StorageAPI) {
// It is okay to have readQuorum disks.
return xl.getLoadBalancedDisks()[0 : xl.readQuorum-1]
}
// getLoadBalancedDisks - fetches load balanced (sufficiently
// randomized) disk slice.
// getLoadBalancedDisks - fetches load balanced (sufficiently randomized) disk slice.
func (xl xlObjects) getLoadBalancedDisks() (disks []StorageAPI) {
// Based on the random shuffling return back randomized disks.
for _, i := range hashOrder(time.Now().UTC().String(), len(xl.storageDisks)) {
@ -60,7 +52,7 @@ func (xl xlObjects) parentDirIsObject(bucket, parent string) bool {
// isObject - returns `true` if the prefix is an object i.e if
// `xl.json` exists at the leaf, false otherwise.
func (xl xlObjects) isObject(bucket, prefix string) (ok bool) {
for _, disk := range xl.getLoadBalancedQuorumDisks() {
for _, disk := range xl.getLoadBalancedDisks() {
if disk == nil {
continue
}

@ -30,7 +30,7 @@ func (xl xlObjects) listObjects(bucket, prefix, marker, delimiter string, maxKey
if walkResultCh == nil {
endWalkCh = make(chan struct{})
isLeaf := xl.isObject
listDir := listDirFactory(isLeaf, xl.getLoadBalancedQuorumDisks()...)
listDir := listDirFactory(isLeaf, xl.getLoadBalancedDisks()...)
walkResultCh = startTreeWalk(bucket, prefix, marker, recursive, listDir, isLeaf, endWalkCh)
}

@ -230,7 +230,7 @@ var objMetadataOpIgnoredErrs = []error{
// readXLMetadata - returns the object metadata `xl.json` content from
// one of the disks picked at random.
func (xl xlObjects) readXLMetadata(bucket, object string) (xlMeta xlMetaV1, err error) {
for _, disk := range xl.getLoadBalancedQuorumDisks() {
for _, disk := range xl.getLoadBalancedDisks() {
if disk == nil {
continue
}

@ -181,7 +181,7 @@ func (xl xlObjects) writeUploadJSON(bucket, object, uploadID string, initiated t
// Returns if the prefix is a multipart upload.
func (xl xlObjects) isMultipartUpload(bucket, prefix string) bool {
for _, disk := range xl.getLoadBalancedQuorumDisks() {
for _, disk := range xl.getLoadBalancedDisks() {
if disk == nil {
continue
}
@ -200,7 +200,7 @@ func (xl xlObjects) isMultipartUpload(bucket, prefix string) bool {
// listUploadsInfo - list all uploads info.
func (xl xlObjects) listUploadsInfo(prefixPath string) (uploadsInfo []uploadInfo, err error) {
for _, disk := range xl.getLoadBalancedQuorumDisks() {
for _, disk := range xl.getLoadBalancedDisks() {
if disk == nil {
continue
}
@ -252,7 +252,7 @@ func (xl xlObjects) removeObjectPart(bucket, object, uploadID, partName string)
// statPart - returns fileInfo structure for a successful stat on part file.
func (xl xlObjects) statPart(bucket, object, uploadID, partName string) (fileInfo FileInfo, err error) {
partNamePath := path.Join(mpartMetaPrefix, bucket, object, uploadID, partName)
for _, disk := range xl.getLoadBalancedQuorumDisks() {
for _, disk := range xl.getLoadBalancedDisks() {
if disk == nil {
continue
}

@ -64,7 +64,7 @@ func (xl xlObjects) listMultipartUploads(bucket, prefix, keyMarker, uploadIDMark
// uploadIDMarker first.
if uploadIDMarker != "" {
nsMutex.RLock(minioMetaBucket, pathJoin(mpartMetaPrefix, bucket, keyMarker))
for _, disk := range xl.getLoadBalancedQuorumDisks() {
for _, disk := range xl.getLoadBalancedDisks() {
if disk == nil {
continue
}
@ -91,7 +91,7 @@ func (xl xlObjects) listMultipartUploads(bucket, prefix, keyMarker, uploadIDMark
if walkerCh == nil {
walkerDoneCh = make(chan struct{})
isLeaf := xl.isMultipartUpload
listDir := listDirFactory(isLeaf, xl.getLoadBalancedQuorumDisks()...)
listDir := listDirFactory(isLeaf, xl.getLoadBalancedDisks()...)
walkerCh = startTreeWalk(minioMetaBucket, multipartPrefixPath, multipartMarkerPath, recursive, listDir, isLeaf, walkerDoneCh)
}
// Collect uploads until we have reached maxUploads count to 0.
@ -130,7 +130,7 @@ func (xl xlObjects) listMultipartUploads(bucket, prefix, keyMarker, uploadIDMark
// For the new object entry we get all its pending uploadIDs.
nsMutex.RLock(minioMetaBucket, pathJoin(mpartMetaPrefix, bucket, entry))
var disk StorageAPI
for _, disk = range xl.getLoadBalancedQuorumDisks() {
for _, disk = range xl.getLoadBalancedDisks() {
if disk == nil {
continue
}
@ -757,7 +757,7 @@ func (xl xlObjects) CompleteMultipartUpload(bucket string, object string, upload
// the object, if yes do not attempt to delete 'uploads.json'.
var disk StorageAPI
var uploadsJSON uploadsV1
for _, disk = range xl.getLoadBalancedQuorumDisks() {
for _, disk = range xl.getLoadBalancedDisks() {
if disk == nil {
continue
}
@ -811,7 +811,7 @@ func (xl xlObjects) abortMultipartUpload(bucket, object, uploadID string) (err e
// the object, if yes do not attempt to delete 'uploads.json'.
var disk StorageAPI
var uploadsJSON uploadsV1
for _, disk = range xl.getLoadBalancedQuorumDisks() {
for _, disk = range xl.getLoadBalancedDisks() {
if disk == nil {
continue
}

Loading…
Cancel
Save