2016-05-21 05:48:47 +02:00
|
|
|
/*
|
|
|
|
* 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.
|
|
|
|
*/
|
|
|
|
|
2016-08-19 01:23:42 +02:00
|
|
|
package cmd
|
2016-05-21 05:48:47 +02:00
|
|
|
|
|
|
|
import (
|
|
|
|
"fmt"
|
2016-05-26 23:13:10 +02:00
|
|
|
"sort"
|
|
|
|
|
2016-08-26 09:11:53 +02:00
|
|
|
"github.com/minio/minio-go/pkg/set"
|
2016-05-26 23:13:10 +02:00
|
|
|
"github.com/minio/minio/pkg/disk"
|
2016-07-04 01:58:21 +02:00
|
|
|
"github.com/minio/minio/pkg/objcache"
|
2016-05-21 05:48:47 +02:00
|
|
|
)
|
|
|
|
|
2016-05-31 01:51:59 +02:00
|
|
|
// XL constants.
|
2016-05-21 05:48:47 +02:00
|
|
|
const (
|
2016-05-31 01:51:59 +02:00
|
|
|
// Format config file carries backend format specific details.
|
2016-05-21 05:48:47 +02:00
|
|
|
formatConfigFile = "format.json"
|
2016-06-23 02:18:31 +02:00
|
|
|
|
|
|
|
// Format config tmp file carries backend format.
|
|
|
|
formatConfigFileTmp = "format.json.tmp"
|
|
|
|
|
2016-05-31 01:51:59 +02:00
|
|
|
// XL metadata file carries per object metadata.
|
|
|
|
xlMetaJSONFile = "xl.json"
|
2016-06-23 02:18:31 +02:00
|
|
|
|
2016-05-31 01:51:59 +02:00
|
|
|
// Uploads metadata file carries per multipart object metadata.
|
|
|
|
uploadsJSONFile = "uploads.json"
|
2016-07-04 01:58:21 +02:00
|
|
|
|
|
|
|
// 8GiB cache by default.
|
|
|
|
maxCacheSize = 8 * 1024 * 1024 * 1024
|
|
|
|
|
|
|
|
// Maximum erasure blocks.
|
|
|
|
maxErasureBlocks = 16
|
|
|
|
|
|
|
|
// Minimum erasure blocks.
|
2016-07-20 04:24:32 +02:00
|
|
|
minErasureBlocks = 4
|
2016-05-21 05:48:47 +02:00
|
|
|
)
|
|
|
|
|
2016-05-31 01:51:59 +02:00
|
|
|
// xlObjects - Implements XL object layer.
|
2016-05-21 05:48:47 +02:00
|
|
|
type xlObjects struct {
|
2016-08-26 02:16:34 +02:00
|
|
|
storageDisks []StorageAPI // Collection of initialized backend disks.
|
|
|
|
dataBlocks int // dataBlocks count caculated for erasure.
|
|
|
|
parityBlocks int // parityBlocks count calculated for erasure.
|
|
|
|
readQuorum int // readQuorum minimum required disks to read data.
|
|
|
|
writeQuorum int // writeQuorum minimum required disks to write data.
|
2016-05-31 01:51:59 +02:00
|
|
|
|
2016-07-04 01:58:21 +02:00
|
|
|
// ListObjects pool management.
|
2016-06-05 20:55:45 +02:00
|
|
|
listPool *treeWalkPool
|
2016-05-21 05:48:47 +02:00
|
|
|
|
2016-07-04 01:58:21 +02:00
|
|
|
// Object cache for caching objects.
|
|
|
|
objCache *objcache.Cache
|
2016-05-21 05:48:47 +02:00
|
|
|
|
2016-07-04 01:58:21 +02:00
|
|
|
// Object cache enabled.
|
|
|
|
objCacheEnabled bool
|
|
|
|
}
|
2016-05-21 05:48:47 +02:00
|
|
|
|
2016-09-15 22:43:40 +02:00
|
|
|
// list of all errors that can be ignored in tree walk operation in XL
|
|
|
|
var xlTreeWalkIgnoredErrs = []error{
|
|
|
|
errFileNotFound,
|
|
|
|
errVolumeNotFound,
|
|
|
|
errDiskNotFound,
|
|
|
|
errDiskAccessDenied,
|
|
|
|
errFaultyDisk,
|
|
|
|
}
|
|
|
|
|
2016-08-29 04:31:59 +02:00
|
|
|
func repairDiskMetadata(storageDisks []StorageAPI) error {
|
2016-06-14 10:12:15 +02:00
|
|
|
// Attempt to load all `format.json`.
|
2016-06-03 01:34:15 +02:00
|
|
|
formatConfigs, sErrs := loadAllFormats(storageDisks)
|
|
|
|
|
2016-06-29 12:13:44 +02:00
|
|
|
// Generic format check validates
|
|
|
|
// if (no quorum) return error
|
|
|
|
// if (disks not recognized) // Always error.
|
2016-06-03 01:34:15 +02:00
|
|
|
if err := genericFormatCheck(formatConfigs, sErrs); err != nil {
|
2016-08-29 04:31:59 +02:00
|
|
|
return err
|
2016-06-03 01:34:15 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
// Handles different cases properly.
|
|
|
|
switch reduceFormatErrs(sErrs, len(storageDisks)) {
|
2016-07-29 01:49:59 +02:00
|
|
|
case errCorruptedFormat:
|
|
|
|
if err := healFormatXLCorruptedDisks(storageDisks); err != nil {
|
2016-08-29 04:31:59 +02:00
|
|
|
return fmt.Errorf("Unable to repair corrupted format, %s", err)
|
2016-07-29 01:49:59 +02:00
|
|
|
}
|
2016-06-03 01:34:15 +02:00
|
|
|
case errSomeDiskUnformatted:
|
|
|
|
// All drives online but some report missing format.json.
|
2016-07-26 12:18:47 +02:00
|
|
|
if err := healFormatXLFreshDisks(storageDisks); err != nil {
|
2016-06-03 01:34:15 +02:00
|
|
|
// There was an unexpected unrecoverable error during healing.
|
2016-08-29 04:31:59 +02:00
|
|
|
return fmt.Errorf("Unable to heal backend %s", err)
|
2016-06-03 01:34:15 +02:00
|
|
|
}
|
|
|
|
case errSomeDiskOffline:
|
2016-07-10 23:38:15 +02:00
|
|
|
// FIXME: in future.
|
2016-08-29 04:31:59 +02:00
|
|
|
return fmt.Errorf("Unable to initialize format %s and %s", errSomeDiskOffline, errSomeDiskUnformatted)
|
|
|
|
}
|
|
|
|
return nil
|
|
|
|
}
|
|
|
|
|
|
|
|
// newXLObjects - initialize new xl object layer.
|
|
|
|
func newXLObjects(disks, ignoredDisks []string) (ObjectLayer, error) {
|
|
|
|
if disks == nil {
|
|
|
|
return nil, errInvalidArgument
|
2016-06-03 01:34:15 +02:00
|
|
|
}
|
2016-08-29 04:31:59 +02:00
|
|
|
disksSet := set.NewStringSet()
|
|
|
|
if len(ignoredDisks) > 0 {
|
|
|
|
disksSet = set.CreateStringSet(ignoredDisks...)
|
|
|
|
}
|
|
|
|
// Bootstrap disks.
|
|
|
|
storageDisks := make([]StorageAPI, len(disks))
|
|
|
|
for index, disk := range disks {
|
|
|
|
// Check if disk is ignored.
|
|
|
|
if disksSet.Contains(disk) {
|
|
|
|
storageDisks[index] = nil
|
|
|
|
continue
|
|
|
|
}
|
|
|
|
var err error
|
|
|
|
// Intentionally ignore disk not found errors. XL is designed
|
|
|
|
// to handle these errors internally.
|
|
|
|
storageDisks[index], err = newStorageAPI(disk)
|
|
|
|
if err != nil && err != errDiskNotFound {
|
|
|
|
switch diskType := storageDisks[index].(type) {
|
|
|
|
case networkStorage:
|
|
|
|
diskType.rpcClient.Close()
|
|
|
|
}
|
|
|
|
return nil, err
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
// Fix format files in case of fresh or corrupted disks
|
|
|
|
repairDiskMetadata(storageDisks)
|
2016-06-03 01:34:15 +02:00
|
|
|
|
2016-07-02 10:59:28 +02:00
|
|
|
// Runs house keeping code, like t, cleaning up tmp files etc.
|
|
|
|
if err := xlHouseKeeping(storageDisks); err != nil {
|
|
|
|
return nil, err
|
|
|
|
}
|
|
|
|
|
2016-05-21 05:48:47 +02:00
|
|
|
// Load saved XL format.json and validate.
|
|
|
|
newPosixDisks, err := loadFormatXL(storageDisks)
|
|
|
|
if err != nil {
|
2016-06-03 01:34:15 +02:00
|
|
|
// errCorruptedDisk - healing failed
|
|
|
|
return nil, fmt.Errorf("Unable to recognize backend format, %s", err)
|
2016-05-21 05:48:47 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
// Calculate data and parity blocks.
|
|
|
|
dataBlocks, parityBlocks := len(newPosixDisks)/2, len(newPosixDisks)/2
|
|
|
|
|
2016-07-09 05:34:27 +02:00
|
|
|
// Initialize object cache.
|
|
|
|
objCache := objcache.New(globalMaxCacheSize, globalCacheExpiry)
|
|
|
|
|
|
|
|
// Initialize list pool.
|
|
|
|
listPool := newTreeWalkPool(globalLookupTimeout)
|
|
|
|
|
2016-06-02 01:43:31 +02:00
|
|
|
// Initialize xl objects.
|
2016-05-21 05:48:47 +02:00
|
|
|
xl := xlObjects{
|
2016-07-09 05:34:27 +02:00
|
|
|
storageDisks: newPosixDisks,
|
|
|
|
dataBlocks: dataBlocks,
|
|
|
|
parityBlocks: parityBlocks,
|
|
|
|
listPool: listPool,
|
|
|
|
objCache: objCache,
|
2016-07-04 01:58:21 +02:00
|
|
|
objCacheEnabled: globalMaxCacheSize > 0,
|
2016-05-21 05:48:47 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
// Figure out read and write quorum based on number of storage disks.
|
2016-07-20 04:24:32 +02:00
|
|
|
// READ and WRITE quorum is always set to (N/2) number of disks.
|
|
|
|
xl.readQuorum = len(xl.storageDisks) / 2
|
2016-07-28 11:20:34 +02:00
|
|
|
xl.writeQuorum = len(xl.storageDisks)/2 + 1
|
2016-05-21 05:48:47 +02:00
|
|
|
|
|
|
|
// Return successfully initialized object layer.
|
|
|
|
return xl, nil
|
|
|
|
}
|
2016-05-26 23:13:10 +02:00
|
|
|
|
2016-08-15 08:55:48 +02:00
|
|
|
// Shutdown function for object storage interface.
|
|
|
|
func (xl xlObjects) Shutdown() error {
|
|
|
|
// Add any object layer shutdown activities here.
|
|
|
|
return nil
|
|
|
|
}
|
|
|
|
|
2016-08-29 04:31:59 +02:00
|
|
|
// HealDiskMetadata function for object storage interface.
|
|
|
|
func (xl xlObjects) HealDiskMetadata() error {
|
2016-08-31 20:39:08 +02:00
|
|
|
// generates random string on setting MINIO_DEBUG=lock, else returns empty string.
|
|
|
|
// used for instrumentation on locks.
|
|
|
|
opsID := getOpsID()
|
|
|
|
|
|
|
|
nsMutex.Lock(minioMetaBucket, formatConfigFile, opsID)
|
|
|
|
defer nsMutex.Unlock(minioMetaBucket, formatConfigFile, opsID)
|
2016-08-29 04:31:59 +02:00
|
|
|
return repairDiskMetadata(xl.storageDisks)
|
|
|
|
}
|
|
|
|
|
2016-05-26 23:13:10 +02:00
|
|
|
// byDiskTotal is a collection satisfying sort.Interface.
|
|
|
|
type byDiskTotal []disk.Info
|
|
|
|
|
|
|
|
func (d byDiskTotal) Len() int { return len(d) }
|
|
|
|
func (d byDiskTotal) Swap(i, j int) { d[i], d[j] = d[j], d[i] }
|
|
|
|
func (d byDiskTotal) Less(i, j int) bool {
|
|
|
|
return d[i].Total < d[j].Total
|
|
|
|
}
|
|
|
|
|
|
|
|
// StorageInfo - returns underlying storage statistics.
|
|
|
|
func (xl xlObjects) StorageInfo() StorageInfo {
|
|
|
|
var disksInfo []disk.Info
|
2016-08-26 02:16:34 +02:00
|
|
|
for _, storageDisk := range xl.storageDisks {
|
2016-09-15 08:53:42 +02:00
|
|
|
if storageDisk == nil {
|
|
|
|
// Storage disk is empty, perhaps ignored disk or not available.
|
|
|
|
continue
|
|
|
|
}
|
2016-08-26 02:16:34 +02:00
|
|
|
info, err := storageDisk.DiskInfo()
|
2016-05-26 23:13:10 +02:00
|
|
|
if err != nil {
|
2016-08-26 02:16:34 +02:00
|
|
|
errorIf(err, "Unable to fetch disk info for %#v", storageDisk)
|
2016-05-26 23:13:10 +02:00
|
|
|
continue
|
|
|
|
}
|
|
|
|
disksInfo = append(disksInfo, info)
|
|
|
|
}
|
|
|
|
|
|
|
|
// Sort so that the first element is the smallest.
|
|
|
|
sort.Sort(byDiskTotal(disksInfo))
|
|
|
|
|
|
|
|
// Return calculated storage info, choose the lowest Total and
|
|
|
|
// Free as the total aggregated values. Total capacity is always
|
|
|
|
// the multiple of smallest disk among the disk list.
|
|
|
|
return StorageInfo{
|
|
|
|
Total: disksInfo[0].Total * int64(len(xl.storageDisks)),
|
|
|
|
Free: disksInfo[0].Free * int64(len(xl.storageDisks)),
|
|
|
|
}
|
|
|
|
}
|