| 
									
										
										
										
											2016-05-31 07:51:59 +08:00
										 |  |  | /* | 
					
						
							|  |  |  |  * Minio Cloud Storage, (C) 2016 Minio, Inc. | 
					
						
							|  |  |  |  * | 
					
						
							|  |  |  |  * Licensed under the Apache License, Version 2.0 (the "License"); | 
					
						
							|  |  |  |  * you may not use this file except in compliance with the License. | 
					
						
							|  |  |  |  * You may obtain a copy of the License at | 
					
						
							|  |  |  |  * | 
					
						
							|  |  |  |  *     http://www.apache.org/licenses/LICENSE-2.0
 | 
					
						
							|  |  |  |  * | 
					
						
							|  |  |  |  * Unless required by applicable law or agreed to in writing, software | 
					
						
							|  |  |  |  * distributed under the License is distributed on an "AS IS" BASIS, | 
					
						
							|  |  |  |  * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. | 
					
						
							|  |  |  |  * See the License for the specific language governing permissions and | 
					
						
							|  |  |  |  * limitations under the License. | 
					
						
							|  |  |  |  */ | 
					
						
							|  |  |  | 
 | 
					
						
							| 
									
										
										
										
											2016-05-26 07:42:31 +08:00
										 |  |  | package main | 
					
						
							|  |  |  | 
 | 
					
						
							|  |  |  | import ( | 
					
						
							| 
									
										
										
										
											2016-06-24 17:06:23 +08:00
										 |  |  | 	"encoding/json" | 
					
						
							| 
									
										
										
										
											2016-07-08 22:33:21 +08:00
										 |  |  | 	"hash/crc32" | 
					
						
							| 
									
										
										
										
											2016-06-03 07:34:15 +08:00
										 |  |  | 	"path" | 
					
						
							| 
									
										
										
										
											2016-07-27 02:34:48 +08:00
										 |  |  | 	"sync" | 
					
						
							| 
									
										
										
										
											2016-05-26 07:42:31 +08:00
										 |  |  | ) | 
					
						
							|  |  |  | 
 | 
					
						
							| 
									
										
										
										
											2016-07-14 02:56:25 +08:00
										 |  |  | // Returns number of errors that occurred the most (incl. nil) and the
 | 
					
						
							|  |  |  | // corresponding error value. N B when there is more than one error value that
 | 
					
						
							|  |  |  | // occurs maximum number of times, the error value returned depends on how
 | 
					
						
							|  |  |  | // golang's map orders keys. This doesn't affect correctness as long as quorum
 | 
					
						
							|  |  |  | // value is greater than or equal to simple majority, since none of the equally
 | 
					
						
							|  |  |  | // maximal values would occur quorum or more number of times.
 | 
					
						
							| 
									
										
										
										
											2016-07-10 04:01:32 +08:00
										 |  |  | 
 | 
					
						
							| 
									
										
										
										
											2016-07-20 10:24:32 +08:00
										 |  |  | func reduceErrs(errs []error, ignoredErrs []error) error { | 
					
						
							| 
									
										
										
										
											2016-07-14 02:56:25 +08:00
										 |  |  | 	errorCounts := make(map[error]int) | 
					
						
							| 
									
										
										
										
											2016-07-10 04:01:32 +08:00
										 |  |  | 	for _, err := range errs { | 
					
						
							| 
									
										
										
										
											2016-07-20 10:24:32 +08:00
										 |  |  | 		if isErrIgnored(err, ignoredErrs) { | 
					
						
							|  |  |  | 			continue | 
					
						
							|  |  |  | 		} | 
					
						
							| 
									
										
										
										
											2016-07-14 02:56:25 +08:00
										 |  |  | 		errorCounts[err]++ | 
					
						
							| 
									
										
										
										
											2016-07-10 04:01:32 +08:00
										 |  |  | 	} | 
					
						
							|  |  |  | 	max := 0 | 
					
						
							| 
									
										
										
										
											2016-07-14 02:56:25 +08:00
										 |  |  | 	var errMax error | 
					
						
							|  |  |  | 	for err, count := range errorCounts { | 
					
						
							|  |  |  | 		if max < count { | 
					
						
							|  |  |  | 			max = count | 
					
						
							|  |  |  | 			errMax = err | 
					
						
							| 
									
										
										
										
											2016-07-10 04:01:32 +08:00
										 |  |  | 		} | 
					
						
							|  |  |  | 	} | 
					
						
							| 
									
										
										
										
											2016-07-20 10:24:32 +08:00
										 |  |  | 	return errMax | 
					
						
							| 
									
										
										
										
											2016-07-10 04:01:32 +08:00
										 |  |  | } | 
					
						
							|  |  |  | 
 | 
					
						
							| 
									
										
										
										
											2016-07-13 15:29:48 +08:00
										 |  |  | // Validates if we have quorum based on the errors related to disk only.
 | 
					
						
							|  |  |  | // Returns 'true' if we have quorum, 'false' if we don't.
 | 
					
						
							|  |  |  | func isDiskQuorum(errs []error, minQuorumCount int) bool { | 
					
						
							|  |  |  | 	var count int | 
					
						
							| 
									
										
										
										
											2016-06-18 02:57:51 +08:00
										 |  |  | 	for _, err := range errs { | 
					
						
							| 
									
										
										
										
											2016-07-13 15:29:48 +08:00
										 |  |  | 		switch err { | 
					
						
							|  |  |  | 		case errDiskNotFound, errFaultyDisk, errDiskAccessDenied: | 
					
						
							| 
									
										
										
										
											2016-06-18 02:57:51 +08:00
										 |  |  | 			continue | 
					
						
							|  |  |  | 		} | 
					
						
							| 
									
										
										
										
											2016-07-13 15:29:48 +08:00
										 |  |  | 		count++ | 
					
						
							| 
									
										
										
										
											2016-06-18 02:57:51 +08:00
										 |  |  | 	} | 
					
						
							| 
									
										
										
										
											2016-07-13 15:29:48 +08:00
										 |  |  | 	return count >= minQuorumCount | 
					
						
							| 
									
										
										
										
											2016-06-18 02:57:51 +08:00
										 |  |  | } | 
					
						
							|  |  |  | 
 | 
					
						
							|  |  |  | // Similar to 'len(slice)' but returns  the actual elements count
 | 
					
						
							|  |  |  | // skipping the unallocated elements.
 | 
					
						
							|  |  |  | func diskCount(disks []StorageAPI) int { | 
					
						
							|  |  |  | 	diskCount := 0 | 
					
						
							|  |  |  | 	for _, disk := range disks { | 
					
						
							|  |  |  | 		if disk == nil { | 
					
						
							|  |  |  | 			continue | 
					
						
							|  |  |  | 		} | 
					
						
							|  |  |  | 		diskCount++ | 
					
						
							|  |  |  | 	} | 
					
						
							|  |  |  | 	return diskCount | 
					
						
							|  |  |  | } | 
					
						
							|  |  |  | 
 | 
					
						
							| 
									
										
										
										
											2016-07-22 10:07:00 +08:00
										 |  |  | // hashOrder - hashes input key to return returns consistent
 | 
					
						
							|  |  |  | // hashed integer slice. Returned integer order is salted
 | 
					
						
							|  |  |  | // with an input key. This results in consistent order.
 | 
					
						
							|  |  |  | // NOTE: collisions are fine, we are not looking for uniqueness
 | 
					
						
							|  |  |  | // in the slices returned.
 | 
					
						
							|  |  |  | func hashOrder(key string, cardinality int) []int { | 
					
						
							|  |  |  | 	if cardinality < 0 { | 
					
						
							|  |  |  | 		// Returns an empty int slice for negative cardinality.
 | 
					
						
							|  |  |  | 		return nil | 
					
						
							| 
									
										
										
										
											2016-05-26 07:42:31 +08:00
										 |  |  | 	} | 
					
						
							| 
									
										
										
										
											2016-07-22 10:07:00 +08:00
										 |  |  | 	nums := make([]int, cardinality) | 
					
						
							|  |  |  | 	keyCrc := crc32.Checksum([]byte(key), crc32.IEEETable) | 
					
						
							| 
									
										
										
										
											2016-07-08 22:33:21 +08:00
										 |  |  | 
 | 
					
						
							| 
									
										
										
										
											2016-07-22 10:07:00 +08:00
										 |  |  | 	start := int(uint32(keyCrc)%uint32(cardinality)) | 1 | 
					
						
							|  |  |  | 	for i := 1; i <= cardinality; i++ { | 
					
						
							|  |  |  | 		nums[i-1] = 1 + ((start + i) % cardinality) | 
					
						
							| 
									
										
										
										
											2016-05-26 07:42:31 +08:00
										 |  |  | 	} | 
					
						
							| 
									
										
										
										
											2016-07-08 22:33:21 +08:00
										 |  |  | 	return nums | 
					
						
							| 
									
										
										
										
											2016-05-26 07:42:31 +08:00
										 |  |  | } | 
					
						
							|  |  |  | 
 | 
					
						
							| 
									
										
										
										
											2016-06-26 05:51:06 +08:00
										 |  |  | // readXLMeta reads `xl.json` and returns back XL metadata structure.
 | 
					
						
							| 
									
										
										
										
											2016-06-24 17:06:23 +08:00
										 |  |  | func readXLMeta(disk StorageAPI, bucket string, object string) (xlMeta xlMetaV1, err error) { | 
					
						
							|  |  |  | 	// Reads entire `xl.json`.
 | 
					
						
							| 
									
										
										
										
											2016-06-26 05:51:06 +08:00
										 |  |  | 	buf, err := disk.ReadAll(bucket, path.Join(object, xlMetaJSONFile)) | 
					
						
							|  |  |  | 	if err != nil { | 
					
						
							| 
									
										
										
										
											2016-06-24 17:06:23 +08:00
										 |  |  | 		return xlMetaV1{}, err | 
					
						
							| 
									
										
										
										
											2016-06-03 07:34:15 +08:00
										 |  |  | 	} | 
					
						
							|  |  |  | 
 | 
					
						
							| 
									
										
										
										
											2016-06-24 17:06:23 +08:00
										 |  |  | 	// Unmarshal xl metadata.
 | 
					
						
							| 
									
										
										
										
											2016-06-26 05:51:06 +08:00
										 |  |  | 	if err = json.Unmarshal(buf, &xlMeta); err != nil { | 
					
						
							| 
									
										
										
										
											2016-06-24 17:06:23 +08:00
										 |  |  | 		return xlMetaV1{}, err | 
					
						
							| 
									
										
										
										
											2016-05-26 07:42:31 +08:00
										 |  |  | 	} | 
					
						
							| 
									
										
										
										
											2016-06-24 17:06:23 +08:00
										 |  |  | 
 | 
					
						
							|  |  |  | 	// Return structured `xl.json`.
 | 
					
						
							|  |  |  | 	return xlMeta, nil | 
					
						
							| 
									
										
										
										
											2016-05-26 07:42:31 +08:00
										 |  |  | } | 
					
						
							| 
									
										
										
										
											2016-07-13 09:23:40 +08:00
										 |  |  | 
 | 
					
						
							| 
									
										
										
										
											2016-07-27 02:34:48 +08:00
										 |  |  | // Reads all `xl.json` metadata as a xlMetaV1 slice.
 | 
					
						
							|  |  |  | // Returns error slice indicating the failed metadata reads.
 | 
					
						
							|  |  |  | func readAllXLMetadata(disks []StorageAPI, bucket, object string) ([]xlMetaV1, []error) { | 
					
						
							|  |  |  | 	errs := make([]error, len(disks)) | 
					
						
							|  |  |  | 	metadataArray := make([]xlMetaV1, len(disks)) | 
					
						
							|  |  |  | 	var wg = &sync.WaitGroup{} | 
					
						
							|  |  |  | 	// Read `xl.json` parallelly across disks.
 | 
					
						
							|  |  |  | 	for index, disk := range disks { | 
					
						
							|  |  |  | 		if disk == nil { | 
					
						
							|  |  |  | 			errs[index] = errDiskNotFound | 
					
						
							|  |  |  | 			continue | 
					
						
							|  |  |  | 		} | 
					
						
							|  |  |  | 		wg.Add(1) | 
					
						
							|  |  |  | 		// Read `xl.json` in routine.
 | 
					
						
							|  |  |  | 		go func(index int, disk StorageAPI) { | 
					
						
							|  |  |  | 			defer wg.Done() | 
					
						
							|  |  |  | 			var err error | 
					
						
							|  |  |  | 			metadataArray[index], err = readXLMeta(disk, bucket, object) | 
					
						
							|  |  |  | 			if err != nil { | 
					
						
							|  |  |  | 				errs[index] = err | 
					
						
							|  |  |  | 				return | 
					
						
							|  |  |  | 			} | 
					
						
							|  |  |  | 		}(index, disk) | 
					
						
							|  |  |  | 	} | 
					
						
							|  |  |  | 
 | 
					
						
							|  |  |  | 	// Wait for all the routines to finish.
 | 
					
						
							|  |  |  | 	wg.Wait() | 
					
						
							|  |  |  | 
 | 
					
						
							|  |  |  | 	// Return all the metadata.
 | 
					
						
							|  |  |  | 	return metadataArray, errs | 
					
						
							|  |  |  | } | 
					
						
							|  |  |  | 
 | 
					
						
							| 
									
										
										
										
											2016-07-16 23:35:30 +08:00
										 |  |  | // Return ordered partsMetadata depeinding on distribution.
 | 
					
						
							|  |  |  | func getOrderedPartsMetadata(distribution []int, partsMetadata []xlMetaV1) (orderedPartsMetadata []xlMetaV1) { | 
					
						
							|  |  |  | 	orderedPartsMetadata = make([]xlMetaV1, len(partsMetadata)) | 
					
						
							|  |  |  | 	for index := range partsMetadata { | 
					
						
							|  |  |  | 		blockIndex := distribution[index] | 
					
						
							|  |  |  | 		orderedPartsMetadata[blockIndex-1] = partsMetadata[index] | 
					
						
							|  |  |  | 	} | 
					
						
							|  |  |  | 	return orderedPartsMetadata | 
					
						
							|  |  |  | } | 
					
						
							|  |  |  | 
 | 
					
						
							|  |  |  | // getOrderedDisks - get ordered disks from erasure distribution.
 | 
					
						
							|  |  |  | // returns ordered slice of disks from their actual distribution.
 | 
					
						
							|  |  |  | func getOrderedDisks(distribution []int, disks []StorageAPI) (orderedDisks []StorageAPI) { | 
					
						
							|  |  |  | 	orderedDisks = make([]StorageAPI, len(disks)) | 
					
						
							|  |  |  | 	// From disks gets ordered disks.
 | 
					
						
							|  |  |  | 	for index := range disks { | 
					
						
							|  |  |  | 		blockIndex := distribution[index] | 
					
						
							|  |  |  | 		orderedDisks[blockIndex-1] = disks[index] | 
					
						
							|  |  |  | 	} | 
					
						
							|  |  |  | 	return orderedDisks | 
					
						
							|  |  |  | } |