2020-06-12 19:28:21 +02:00
|
|
|
/*
|
|
|
|
* MinIO Cloud Storage, (C) 2020 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.
|
|
|
|
*/
|
|
|
|
|
|
|
|
package cmd
|
|
|
|
|
|
|
|
import (
|
|
|
|
"bytes"
|
|
|
|
"context"
|
|
|
|
"encoding/binary"
|
|
|
|
"errors"
|
2020-09-26 04:21:52 +02:00
|
|
|
"math/rand"
|
2020-06-12 19:28:21 +02:00
|
|
|
"os"
|
|
|
|
"path"
|
|
|
|
"strconv"
|
|
|
|
"strings"
|
|
|
|
"time"
|
|
|
|
|
|
|
|
"github.com/minio/minio/cmd/config"
|
2020-10-14 22:51:51 +02:00
|
|
|
"github.com/minio/minio/cmd/config/heal"
|
2020-06-12 19:28:21 +02:00
|
|
|
"github.com/minio/minio/cmd/logger"
|
|
|
|
"github.com/minio/minio/pkg/bucket/lifecycle"
|
2020-07-22 02:49:56 +02:00
|
|
|
"github.com/minio/minio/pkg/bucket/replication"
|
2020-06-12 19:28:21 +02:00
|
|
|
"github.com/minio/minio/pkg/color"
|
|
|
|
"github.com/minio/minio/pkg/env"
|
|
|
|
"github.com/minio/minio/pkg/event"
|
|
|
|
"github.com/minio/minio/pkg/hash"
|
2020-09-12 09:08:12 +02:00
|
|
|
"github.com/minio/minio/pkg/madmin"
|
2020-06-12 19:28:21 +02:00
|
|
|
"github.com/willf/bloom"
|
|
|
|
)
|
|
|
|
|
|
|
|
const (
|
|
|
|
dataCrawlSleepPerFolder = time.Millisecond // Time to wait between folders.
|
|
|
|
dataCrawlSleepDefMult = 10.0 // Default multiplier for waits between operations.
|
2020-08-24 23:05:12 +02:00
|
|
|
dataCrawlStartDelay = 5 * time.Minute // Time to wait on startup and between cycles.
|
2020-06-12 19:28:21 +02:00
|
|
|
dataUsageUpdateDirCycles = 16 // Visit all folders every n cycles.
|
|
|
|
|
2020-09-12 09:08:12 +02:00
|
|
|
healDeleteDangling = true
|
|
|
|
healFolderIncludeProb = 32 // Include a clean folder one in n cycles.
|
|
|
|
healObjectSelectProb = 512 // Overall probability of a file being scanned; one in n.
|
|
|
|
)
|
|
|
|
|
|
|
|
var (
|
2020-10-14 22:51:51 +02:00
|
|
|
globalHealConfig heal.Config
|
2020-10-08 21:32:32 +02:00
|
|
|
dataCrawlerLeaderLockTimeout = newDynamicTimeout(30*time.Second, 10*time.Second)
|
2020-06-12 19:28:21 +02:00
|
|
|
)
|
|
|
|
|
|
|
|
// initDataCrawler will start the crawler unless disabled.
|
|
|
|
func initDataCrawler(ctx context.Context, objAPI ObjectLayer) {
|
|
|
|
if env.Get(envDataUsageCrawlConf, config.EnableOn) == config.EnableOn {
|
|
|
|
go runDataCrawler(ctx, objAPI)
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
// runDataCrawler will start a data crawler.
|
|
|
|
// The function will block until the context is canceled.
|
|
|
|
// There should only ever be one crawler running per cluster.
|
|
|
|
func runDataCrawler(ctx context.Context, objAPI ObjectLayer) {
|
2020-09-18 20:15:54 +02:00
|
|
|
// Make sure only 1 crawler is running on the cluster.
|
2020-11-04 17:25:42 +01:00
|
|
|
locker := objAPI.NewNSLock(minioMetaBucket, "runDataCrawler.lock")
|
2020-09-26 04:21:52 +02:00
|
|
|
r := rand.New(rand.NewSource(time.Now().UnixNano()))
|
2020-09-18 20:15:54 +02:00
|
|
|
for {
|
2020-11-04 17:25:42 +01:00
|
|
|
err := locker.GetLock(ctx, dataCrawlerLeaderLockTimeout)
|
2020-09-18 20:15:54 +02:00
|
|
|
if err != nil {
|
2020-09-26 04:21:52 +02:00
|
|
|
time.Sleep(time.Duration(r.Float64() * float64(dataCrawlStartDelay)))
|
2020-09-18 20:15:54 +02:00
|
|
|
continue
|
|
|
|
}
|
|
|
|
break
|
|
|
|
// No unlock for "leader" lock.
|
|
|
|
}
|
|
|
|
|
2020-06-12 19:28:21 +02:00
|
|
|
// Load current bloom cycle
|
|
|
|
nextBloomCycle := intDataUpdateTracker.current() + 1
|
|
|
|
var buf bytes.Buffer
|
|
|
|
err := objAPI.GetObject(ctx, dataUsageBucket, dataUsageBloomName, 0, -1, &buf, "", ObjectOptions{})
|
|
|
|
if err != nil {
|
|
|
|
if !isErrObjectNotFound(err) && !isErrBucketNotFound(err) {
|
|
|
|
logger.LogIf(ctx, err)
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
if buf.Len() == 8 {
|
|
|
|
nextBloomCycle = binary.LittleEndian.Uint64(buf.Bytes())
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
for {
|
|
|
|
select {
|
|
|
|
case <-ctx.Done():
|
|
|
|
return
|
|
|
|
case <-time.NewTimer(dataCrawlStartDelay).C:
|
|
|
|
// Wait before starting next cycle and wait on startup.
|
|
|
|
results := make(chan DataUsageInfo, 1)
|
|
|
|
go storeDataUsageInBackend(ctx, objAPI, results)
|
|
|
|
bf, err := globalNotificationSys.updateBloomFilter(ctx, nextBloomCycle)
|
|
|
|
logger.LogIf(ctx, err)
|
|
|
|
err = objAPI.CrawlAndGetDataUsage(ctx, bf, results)
|
|
|
|
close(results)
|
|
|
|
logger.LogIf(ctx, err)
|
|
|
|
if err == nil {
|
|
|
|
// Store new cycle...
|
|
|
|
nextBloomCycle++
|
|
|
|
var tmp [8]byte
|
|
|
|
binary.LittleEndian.PutUint64(tmp[:], nextBloomCycle)
|
|
|
|
r, err := hash.NewReader(bytes.NewReader(tmp[:]), int64(len(tmp)), "", "", int64(len(tmp)), false)
|
|
|
|
if err != nil {
|
|
|
|
logger.LogIf(ctx, err)
|
|
|
|
continue
|
|
|
|
}
|
|
|
|
|
|
|
|
_, err = objAPI.PutObject(ctx, dataUsageBucket, dataUsageBloomName, NewPutObjReader(r, nil, nil), ObjectOptions{})
|
|
|
|
if !isErrBucketNotFound(err) {
|
|
|
|
logger.LogIf(ctx, err)
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
type cachedFolder struct {
|
2020-08-24 22:47:01 +02:00
|
|
|
name string
|
|
|
|
parent *dataUsageHash
|
|
|
|
objectHealProbDiv uint32
|
2020-06-12 19:28:21 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
type folderScanner struct {
|
2020-10-13 22:45:08 +02:00
|
|
|
root string
|
|
|
|
getSize getSizeFn
|
|
|
|
oldCache dataUsageCache
|
|
|
|
newCache dataUsageCache
|
|
|
|
withFilter *bloomFilter
|
2020-06-12 19:28:21 +02:00
|
|
|
|
|
|
|
dataUsageCrawlMult float64
|
|
|
|
dataUsageCrawlDebug bool
|
2020-08-24 22:47:01 +02:00
|
|
|
healFolderInclude uint32 // Include a clean folder one in n cycles.
|
|
|
|
healObjectSelect uint32 // Do a heal check on an object once every n cycles. Must divide into healFolderInclude
|
2020-06-12 19:28:21 +02:00
|
|
|
|
|
|
|
newFolders []cachedFolder
|
|
|
|
existingFolders []cachedFolder
|
|
|
|
}
|
|
|
|
|
|
|
|
// crawlDataFolder will crawl the basepath+cache.Info.Name and return an updated cache.
|
|
|
|
// The returned cache will always be valid, but may not be updated from the existing.
|
|
|
|
// Before each operation waitForLowActiveIO is called which can be used to temporarily halt the crawler.
|
|
|
|
// If the supplied context is canceled the function will return at the first chance.
|
2020-10-13 22:45:08 +02:00
|
|
|
func crawlDataFolder(ctx context.Context, basePath string, cache dataUsageCache, getSize getSizeFn) (dataUsageCache, error) {
|
2020-06-12 19:28:21 +02:00
|
|
|
t := UTCNow()
|
|
|
|
|
|
|
|
logPrefix := color.Green("data-usage: ")
|
|
|
|
logSuffix := color.Blue(" - %v + %v", basePath, cache.Info.Name)
|
|
|
|
if intDataUpdateTracker.debug {
|
|
|
|
defer func() {
|
|
|
|
logger.Info(logPrefix+" Crawl time: %v"+logSuffix, time.Since(t))
|
|
|
|
}()
|
|
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
switch cache.Info.Name {
|
|
|
|
case "", dataUsageRoot:
|
|
|
|
return cache, errors.New("internal error: root scan attempted")
|
|
|
|
}
|
|
|
|
|
|
|
|
delayMult, err := strconv.ParseFloat(env.Get(envDataUsageCrawlDelay, "10.0"), 64)
|
|
|
|
if err != nil {
|
|
|
|
logger.LogIf(ctx, err)
|
|
|
|
delayMult = dataCrawlSleepDefMult
|
|
|
|
}
|
|
|
|
|
|
|
|
s := folderScanner{
|
|
|
|
root: basePath,
|
|
|
|
getSize: getSize,
|
|
|
|
oldCache: cache,
|
|
|
|
newCache: dataUsageCache{Info: cache.Info},
|
|
|
|
newFolders: nil,
|
|
|
|
existingFolders: nil,
|
|
|
|
dataUsageCrawlMult: delayMult,
|
|
|
|
dataUsageCrawlDebug: intDataUpdateTracker.debug,
|
2020-08-24 22:47:01 +02:00
|
|
|
healFolderInclude: 0,
|
|
|
|
healObjectSelect: 0,
|
2020-06-12 19:28:21 +02:00
|
|
|
}
|
|
|
|
|
2020-08-24 22:47:01 +02:00
|
|
|
// Enable healing in XL mode.
|
|
|
|
if globalIsErasure {
|
|
|
|
// Include a clean folder one in n cycles.
|
2020-09-12 09:08:12 +02:00
|
|
|
s.healFolderInclude = healFolderIncludeProb
|
2020-08-24 22:47:01 +02:00
|
|
|
// Do a heal check on an object once every n cycles. Must divide into healFolderInclude
|
2020-09-12 09:08:12 +02:00
|
|
|
s.healObjectSelect = healObjectSelectProb
|
2020-08-24 22:47:01 +02:00
|
|
|
}
|
2020-06-12 19:28:21 +02:00
|
|
|
if len(cache.Info.BloomFilter) > 0 {
|
|
|
|
s.withFilter = &bloomFilter{BloomFilter: &bloom.BloomFilter{}}
|
|
|
|
_, err := s.withFilter.ReadFrom(bytes.NewBuffer(cache.Info.BloomFilter))
|
|
|
|
if err != nil {
|
|
|
|
logger.LogIf(ctx, err, logPrefix+"Error reading bloom filter")
|
|
|
|
s.withFilter = nil
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if s.dataUsageCrawlDebug {
|
|
|
|
logger.Info(logPrefix+"Start crawling. Bloom filter: %v"+logSuffix, s.withFilter != nil)
|
|
|
|
}
|
|
|
|
|
|
|
|
done := ctx.Done()
|
|
|
|
var flattenLevels = 2
|
|
|
|
|
|
|
|
if s.dataUsageCrawlDebug {
|
|
|
|
logger.Info(logPrefix+"Cycle: %v, Entries: %v"+logSuffix, cache.Info.NextCycle, len(cache.Cache))
|
|
|
|
}
|
|
|
|
|
|
|
|
// Always scan flattenLevels deep. Cache root is level 0.
|
2020-08-24 22:47:01 +02:00
|
|
|
todo := []cachedFolder{{name: cache.Info.Name, objectHealProbDiv: 1}}
|
2020-06-12 19:28:21 +02:00
|
|
|
for i := 0; i < flattenLevels; i++ {
|
|
|
|
if s.dataUsageCrawlDebug {
|
|
|
|
logger.Info(logPrefix+"Level %v, scanning %v directories."+logSuffix, i, len(todo))
|
|
|
|
}
|
|
|
|
select {
|
|
|
|
case <-done:
|
|
|
|
return cache, ctx.Err()
|
|
|
|
default:
|
|
|
|
}
|
|
|
|
var err error
|
|
|
|
todo, err = s.scanQueuedLevels(ctx, todo, i == flattenLevels-1)
|
|
|
|
if err != nil {
|
|
|
|
// No useful information...
|
|
|
|
return cache, err
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if s.dataUsageCrawlDebug {
|
|
|
|
logger.Info(logPrefix+"New folders: %v"+logSuffix, s.newFolders)
|
|
|
|
}
|
|
|
|
|
|
|
|
// Add new folders first
|
|
|
|
for _, folder := range s.newFolders {
|
|
|
|
select {
|
|
|
|
case <-done:
|
|
|
|
return s.newCache, ctx.Err()
|
|
|
|
default:
|
|
|
|
}
|
2020-08-24 22:47:01 +02:00
|
|
|
du, err := s.deepScanFolder(ctx, folder)
|
2020-06-12 19:28:21 +02:00
|
|
|
if err != nil {
|
|
|
|
logger.LogIf(ctx, err)
|
|
|
|
continue
|
|
|
|
}
|
|
|
|
if du == nil {
|
|
|
|
logger.Info(logPrefix + "no disk usage provided" + logSuffix)
|
|
|
|
continue
|
|
|
|
}
|
|
|
|
|
|
|
|
s.newCache.replace(folder.name, "", *du)
|
|
|
|
// Add to parent manually
|
|
|
|
if folder.parent != nil {
|
|
|
|
parent := s.newCache.Cache[folder.parent.Key()]
|
|
|
|
parent.addChildString(folder.name)
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if s.dataUsageCrawlDebug {
|
|
|
|
logger.Info(logPrefix+"Existing folders: %v"+logSuffix, len(s.existingFolders))
|
|
|
|
}
|
|
|
|
|
|
|
|
// Do selective scanning of existing folders.
|
|
|
|
for _, folder := range s.existingFolders {
|
|
|
|
select {
|
|
|
|
case <-done:
|
|
|
|
return s.newCache, ctx.Err()
|
|
|
|
default:
|
|
|
|
}
|
|
|
|
h := hashPath(folder.name)
|
|
|
|
if !h.mod(s.oldCache.Info.NextCycle, dataUsageUpdateDirCycles) {
|
2020-08-24 22:47:01 +02:00
|
|
|
if !h.mod(s.oldCache.Info.NextCycle, s.healFolderInclude/folder.objectHealProbDiv) {
|
|
|
|
s.newCache.replaceHashed(h, folder.parent, s.oldCache.Cache[h.Key()])
|
|
|
|
continue
|
|
|
|
} else {
|
|
|
|
folder.objectHealProbDiv = s.healFolderInclude
|
|
|
|
}
|
|
|
|
folder.objectHealProbDiv = dataUsageUpdateDirCycles
|
2020-06-12 19:28:21 +02:00
|
|
|
}
|
|
|
|
if s.withFilter != nil {
|
|
|
|
_, prefix := path2BucketObjectWithBasePath(basePath, folder.name)
|
|
|
|
if s.oldCache.Info.lifeCycle == nil || !s.oldCache.Info.lifeCycle.HasActiveRules(prefix, true) {
|
|
|
|
// If folder isn't in filter, skip it completely.
|
|
|
|
if !s.withFilter.containsDir(folder.name) {
|
2020-08-24 22:47:01 +02:00
|
|
|
if !h.mod(s.oldCache.Info.NextCycle, s.healFolderInclude/folder.objectHealProbDiv) {
|
|
|
|
if s.dataUsageCrawlDebug {
|
|
|
|
logger.Info(logPrefix+"Skipping non-updated folder: %v"+logSuffix, folder)
|
|
|
|
}
|
|
|
|
s.newCache.replaceHashed(h, folder.parent, s.oldCache.Cache[h.Key()])
|
|
|
|
continue
|
|
|
|
} else {
|
|
|
|
if s.dataUsageCrawlDebug {
|
|
|
|
logger.Info(logPrefix+"Adding non-updated folder to heal check: %v"+logSuffix, folder.name)
|
|
|
|
}
|
|
|
|
// Update probability of including objects
|
|
|
|
folder.objectHealProbDiv = s.healFolderInclude
|
2020-06-12 19:28:21 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
// Update on this cycle...
|
2020-08-24 22:47:01 +02:00
|
|
|
du, err := s.deepScanFolder(ctx, folder)
|
2020-06-12 19:28:21 +02:00
|
|
|
if err != nil {
|
|
|
|
logger.LogIf(ctx, err)
|
|
|
|
continue
|
|
|
|
}
|
|
|
|
if du == nil {
|
|
|
|
logger.LogIf(ctx, errors.New("data-usage: no disk usage provided"))
|
|
|
|
continue
|
|
|
|
}
|
|
|
|
s.newCache.replaceHashed(h, folder.parent, *du)
|
|
|
|
}
|
|
|
|
if s.dataUsageCrawlDebug {
|
|
|
|
logger.Info(logPrefix+"Finished crawl, %v entries"+logSuffix, len(s.newCache.Cache))
|
|
|
|
}
|
|
|
|
s.newCache.Info.LastUpdate = UTCNow()
|
|
|
|
s.newCache.Info.NextCycle++
|
|
|
|
return s.newCache, nil
|
|
|
|
}
|
|
|
|
|
|
|
|
// scanQueuedLevels will scan the provided folders.
|
|
|
|
// Files found in the folders will be added to f.newCache.
|
|
|
|
// If final is provided folders will be put into f.newFolders or f.existingFolders.
|
|
|
|
// If final is not provided the folders found are returned from the function.
|
|
|
|
func (f *folderScanner) scanQueuedLevels(ctx context.Context, folders []cachedFolder, final bool) ([]cachedFolder, error) {
|
|
|
|
var nextFolders []cachedFolder
|
|
|
|
done := ctx.Done()
|
|
|
|
for _, folder := range folders {
|
|
|
|
select {
|
|
|
|
case <-done:
|
|
|
|
return nil, ctx.Err()
|
|
|
|
default:
|
|
|
|
}
|
|
|
|
thisHash := hashPath(folder.name)
|
2020-08-24 22:47:01 +02:00
|
|
|
existing := f.oldCache.findChildrenCopy(thisHash)
|
2020-06-12 19:28:21 +02:00
|
|
|
|
|
|
|
// If there are lifecycle rules for the prefix, remove the filter.
|
|
|
|
filter := f.withFilter
|
|
|
|
var activeLifeCycle *lifecycle.Lifecycle
|
2020-09-21 20:41:07 +02:00
|
|
|
if f.oldCache.Info.lifeCycle != nil {
|
2020-06-12 19:28:21 +02:00
|
|
|
_, prefix := path2BucketObjectWithBasePath(f.root, folder.name)
|
|
|
|
if f.oldCache.Info.lifeCycle.HasActiveRules(prefix, true) {
|
|
|
|
if f.dataUsageCrawlDebug {
|
2020-08-24 22:47:01 +02:00
|
|
|
logger.Info(color.Green("folder-scanner:")+" Prefix %q has active rules", prefix)
|
2020-06-12 19:28:21 +02:00
|
|
|
}
|
|
|
|
activeLifeCycle = f.oldCache.Info.lifeCycle
|
|
|
|
filter = nil
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if _, ok := f.oldCache.Cache[thisHash.Key()]; filter != nil && ok {
|
|
|
|
// If folder isn't in filter and we have data, skip it completely.
|
|
|
|
if folder.name != dataUsageRoot && !filter.containsDir(folder.name) {
|
2020-08-24 22:47:01 +02:00
|
|
|
if !thisHash.mod(f.oldCache.Info.NextCycle, f.healFolderInclude/folder.objectHealProbDiv) {
|
|
|
|
f.newCache.copyWithChildren(&f.oldCache, thisHash, folder.parent)
|
|
|
|
if f.dataUsageCrawlDebug {
|
|
|
|
logger.Info(color.Green("folder-scanner:")+" Skipping non-updated folder: %v", folder.name)
|
|
|
|
}
|
|
|
|
continue
|
|
|
|
} else {
|
|
|
|
if f.dataUsageCrawlDebug {
|
|
|
|
logger.Info(color.Green("folder-scanner:")+" Adding non-updated folder to heal check: %v", folder.name)
|
|
|
|
}
|
|
|
|
// If probability was already crawlerHealFolderInclude, keep it.
|
|
|
|
folder.objectHealProbDiv = f.healFolderInclude
|
2020-06-12 19:28:21 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
sleepDuration(dataCrawlSleepPerFolder, f.dataUsageCrawlMult)
|
|
|
|
|
|
|
|
cache := dataUsageEntry{}
|
|
|
|
|
|
|
|
err := readDirFn(path.Join(f.root, folder.name), func(entName string, typ os.FileMode) error {
|
|
|
|
// Parse
|
|
|
|
entName = path.Clean(path.Join(folder.name, entName))
|
|
|
|
bucket, prefix := path2BucketObjectWithBasePath(f.root, entName)
|
|
|
|
if bucket == "" {
|
|
|
|
if f.dataUsageCrawlDebug {
|
2020-08-24 22:47:01 +02:00
|
|
|
logger.Info(color.Green("folder-scanner:")+" no bucket (%s,%s)", f.root, entName)
|
2020-06-12 19:28:21 +02:00
|
|
|
}
|
|
|
|
return nil
|
|
|
|
}
|
|
|
|
|
|
|
|
if isReservedOrInvalidBucket(bucket, false) {
|
|
|
|
if f.dataUsageCrawlDebug {
|
2020-08-24 22:47:01 +02:00
|
|
|
logger.Info(color.Green("folder-scanner:")+" invalid bucket: %v, entry: %v", bucket, entName)
|
2020-06-12 19:28:21 +02:00
|
|
|
}
|
|
|
|
return nil
|
|
|
|
}
|
|
|
|
|
|
|
|
select {
|
|
|
|
case <-done:
|
|
|
|
return ctx.Err()
|
|
|
|
default:
|
|
|
|
}
|
|
|
|
|
|
|
|
if typ&os.ModeDir != 0 {
|
|
|
|
h := hashPath(entName)
|
|
|
|
_, exists := f.oldCache.Cache[h.Key()]
|
|
|
|
cache.addChildString(entName)
|
|
|
|
|
2020-08-24 22:47:01 +02:00
|
|
|
this := cachedFolder{name: entName, parent: &thisHash, objectHealProbDiv: folder.objectHealProbDiv}
|
|
|
|
delete(existing, h.Key())
|
2020-06-12 19:28:21 +02:00
|
|
|
cache.addChild(h)
|
|
|
|
if final {
|
|
|
|
if exists {
|
|
|
|
f.existingFolders = append(f.existingFolders, this)
|
|
|
|
} else {
|
|
|
|
f.newFolders = append(f.newFolders, this)
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
nextFolders = append(nextFolders, this)
|
|
|
|
}
|
|
|
|
return nil
|
|
|
|
}
|
2020-10-22 22:36:24 +02:00
|
|
|
|
2020-06-12 19:28:21 +02:00
|
|
|
// Dynamic time delay.
|
|
|
|
t := UTCNow()
|
|
|
|
|
|
|
|
// Get file size, ignore errors.
|
|
|
|
item := crawlItem{
|
|
|
|
Path: path.Join(f.root, entName),
|
|
|
|
Typ: typ,
|
|
|
|
bucket: bucket,
|
|
|
|
prefix: path.Dir(prefix),
|
|
|
|
objectName: path.Base(entName),
|
|
|
|
debug: f.dataUsageCrawlDebug,
|
|
|
|
lifeCycle: activeLifeCycle,
|
2020-08-24 22:47:01 +02:00
|
|
|
heal: thisHash.mod(f.oldCache.Info.NextCycle, f.healObjectSelect/folder.objectHealProbDiv),
|
2020-06-12 19:28:21 +02:00
|
|
|
}
|
|
|
|
size, err := f.getSize(item)
|
|
|
|
|
|
|
|
sleepDuration(time.Since(t), f.dataUsageCrawlMult)
|
2020-08-24 22:47:01 +02:00
|
|
|
if err == errSkipFile {
|
2020-06-12 19:28:21 +02:00
|
|
|
return nil
|
|
|
|
}
|
|
|
|
logger.LogIf(ctx, err)
|
|
|
|
cache.Size += size
|
|
|
|
cache.Objects++
|
|
|
|
cache.ObjSizes.add(size)
|
|
|
|
|
|
|
|
return nil
|
|
|
|
})
|
|
|
|
if err != nil {
|
|
|
|
return nil, err
|
|
|
|
}
|
2020-08-24 22:47:01 +02:00
|
|
|
|
|
|
|
if f.healObjectSelect == 0 {
|
|
|
|
// If we are not scanning, return now.
|
|
|
|
f.newCache.replaceHashed(thisHash, folder.parent, cache)
|
|
|
|
continue
|
|
|
|
}
|
|
|
|
|
2020-10-09 18:59:52 +02:00
|
|
|
objAPI := newObjectLayerFn()
|
2020-08-24 22:47:01 +02:00
|
|
|
if objAPI == nil {
|
|
|
|
continue
|
|
|
|
}
|
|
|
|
|
|
|
|
bgSeq, found := globalBackgroundHealState.getHealSequenceByToken(bgHealingUUID)
|
|
|
|
if !found {
|
|
|
|
continue
|
|
|
|
}
|
|
|
|
|
|
|
|
// Whatever remains in 'existing' are folders at this level
|
|
|
|
// that existed in the previous run but wasn't found now.
|
|
|
|
//
|
|
|
|
// This may be because of 2 reasons:
|
|
|
|
//
|
|
|
|
// 1) The folder/object was deleted.
|
|
|
|
// 2) We come from another disk and this disk missed the write.
|
|
|
|
//
|
|
|
|
// We therefore perform a heal check.
|
|
|
|
// If that doesn't bring it back we remove the folder and assume it was deleted.
|
|
|
|
// This means that the next run will not look for it.
|
|
|
|
for k := range existing {
|
|
|
|
bucket, prefix := path2BucketObject(k)
|
|
|
|
if f.dataUsageCrawlDebug {
|
|
|
|
logger.Info(color.Green("folder-scanner:")+" checking disappeared folder: %v/%v", bucket, prefix)
|
|
|
|
}
|
|
|
|
|
2020-10-22 22:36:24 +02:00
|
|
|
// Dynamic time delay.
|
|
|
|
t := UTCNow()
|
|
|
|
|
2020-08-24 22:47:01 +02:00
|
|
|
err = objAPI.HealObjects(ctx, bucket, prefix, madmin.HealOpts{Recursive: true, Remove: healDeleteDangling},
|
|
|
|
func(bucket, object, versionID string) error {
|
2020-10-22 22:36:24 +02:00
|
|
|
// Wait for each heal as per crawler frequency.
|
|
|
|
sleepDuration(time.Since(t), f.dataUsageCrawlMult)
|
|
|
|
|
|
|
|
defer func() {
|
|
|
|
t = UTCNow()
|
|
|
|
}()
|
2020-08-24 22:47:01 +02:00
|
|
|
return bgSeq.queueHealTask(healSource{
|
|
|
|
bucket: bucket,
|
|
|
|
object: object,
|
|
|
|
versionID: versionID,
|
|
|
|
}, madmin.HealItemObject)
|
|
|
|
})
|
2020-10-22 22:36:24 +02:00
|
|
|
|
2020-10-13 22:45:08 +02:00
|
|
|
sleepDuration(time.Since(t), f.dataUsageCrawlMult)
|
2020-08-24 22:47:01 +02:00
|
|
|
|
|
|
|
if f.dataUsageCrawlDebug && err != nil {
|
|
|
|
logger.Info(color.Green("healObjects:")+" checking returned value %v", err)
|
|
|
|
}
|
|
|
|
|
|
|
|
// Add unless healing returned an error.
|
|
|
|
if err == nil {
|
|
|
|
this := cachedFolder{name: k, parent: &thisHash, objectHealProbDiv: folder.objectHealProbDiv}
|
|
|
|
cache.addChild(hashPath(k))
|
|
|
|
if final {
|
|
|
|
f.existingFolders = append(f.existingFolders, this)
|
|
|
|
} else {
|
|
|
|
nextFolders = append(nextFolders, this)
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
2020-06-12 19:28:21 +02:00
|
|
|
f.newCache.replaceHashed(thisHash, folder.parent, cache)
|
|
|
|
}
|
|
|
|
return nextFolders, nil
|
|
|
|
}
|
|
|
|
|
|
|
|
// deepScanFolder will deep scan a folder and return the size if no error occurs.
|
2020-08-24 22:47:01 +02:00
|
|
|
func (f *folderScanner) deepScanFolder(ctx context.Context, folder cachedFolder) (*dataUsageEntry, error) {
|
2020-06-12 19:28:21 +02:00
|
|
|
var cache dataUsageEntry
|
|
|
|
|
|
|
|
done := ctx.Done()
|
|
|
|
|
|
|
|
var addDir func(entName string, typ os.FileMode) error
|
2020-08-24 22:47:01 +02:00
|
|
|
var dirStack = []string{f.root, folder.name}
|
2020-06-12 19:28:21 +02:00
|
|
|
|
|
|
|
addDir = func(entName string, typ os.FileMode) error {
|
|
|
|
select {
|
|
|
|
case <-done:
|
|
|
|
return ctx.Err()
|
|
|
|
default:
|
|
|
|
}
|
|
|
|
|
|
|
|
if typ&os.ModeDir != 0 {
|
|
|
|
dirStack = append(dirStack, entName)
|
|
|
|
err := readDirFn(path.Join(dirStack...), addDir)
|
|
|
|
dirStack = dirStack[:len(dirStack)-1]
|
|
|
|
sleepDuration(dataCrawlSleepPerFolder, f.dataUsageCrawlMult)
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
|
|
|
|
// Dynamic time delay.
|
|
|
|
t := UTCNow()
|
|
|
|
|
|
|
|
// Get file size, ignore errors.
|
|
|
|
dirStack = append(dirStack, entName)
|
|
|
|
fileName := path.Join(dirStack...)
|
|
|
|
dirStack = dirStack[:len(dirStack)-1]
|
|
|
|
|
|
|
|
bucket, prefix := path2BucketObjectWithBasePath(f.root, fileName)
|
|
|
|
var activeLifeCycle *lifecycle.Lifecycle
|
|
|
|
if f.oldCache.Info.lifeCycle != nil {
|
|
|
|
if f.oldCache.Info.lifeCycle.HasActiveRules(prefix, false) {
|
|
|
|
if f.dataUsageCrawlDebug {
|
2020-08-24 22:47:01 +02:00
|
|
|
logger.Info(color.Green("folder-scanner:")+" Prefix %q has active rules", prefix)
|
2020-06-12 19:28:21 +02:00
|
|
|
}
|
|
|
|
activeLifeCycle = f.oldCache.Info.lifeCycle
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
size, err := f.getSize(
|
|
|
|
crawlItem{
|
|
|
|
Path: fileName,
|
|
|
|
Typ: typ,
|
|
|
|
bucket: bucket,
|
|
|
|
prefix: path.Dir(prefix),
|
|
|
|
objectName: path.Base(entName),
|
|
|
|
debug: f.dataUsageCrawlDebug,
|
|
|
|
lifeCycle: activeLifeCycle,
|
2020-08-24 22:47:01 +02:00
|
|
|
heal: hashPath(path.Join(prefix, entName)).mod(f.oldCache.Info.NextCycle, f.healObjectSelect/folder.objectHealProbDiv),
|
2020-06-12 19:28:21 +02:00
|
|
|
})
|
|
|
|
|
|
|
|
// Don't sleep for really small amount of time
|
|
|
|
sleepDuration(time.Since(t), f.dataUsageCrawlMult)
|
|
|
|
|
|
|
|
if err == errSkipFile {
|
|
|
|
return nil
|
|
|
|
}
|
|
|
|
logger.LogIf(ctx, err)
|
|
|
|
cache.Size += size
|
|
|
|
cache.Objects++
|
|
|
|
cache.ObjSizes.add(size)
|
|
|
|
return nil
|
|
|
|
}
|
|
|
|
err := readDirFn(path.Join(dirStack...), addDir)
|
|
|
|
if err != nil {
|
|
|
|
return nil, err
|
|
|
|
}
|
|
|
|
return &cache, nil
|
|
|
|
}
|
|
|
|
|
|
|
|
// crawlItem represents each file while walking.
|
|
|
|
type crawlItem struct {
|
|
|
|
Path string
|
|
|
|
Typ os.FileMode
|
|
|
|
|
|
|
|
bucket string // Bucket.
|
|
|
|
prefix string // Only the prefix if any, does not have final object name.
|
|
|
|
objectName string // Only the object name without prefixes.
|
|
|
|
lifeCycle *lifecycle.Lifecycle
|
2020-08-24 22:47:01 +02:00
|
|
|
heal bool // Has the object been selected for heal check?
|
2020-06-12 19:28:21 +02:00
|
|
|
debug bool
|
|
|
|
}
|
|
|
|
|
|
|
|
type getSizeFn func(item crawlItem) (int64, error)
|
|
|
|
|
|
|
|
// transformMetaDir will transform a directory to prefix/file.ext
|
|
|
|
func (i *crawlItem) transformMetaDir() {
|
|
|
|
split := strings.Split(i.prefix, SlashSeparator)
|
|
|
|
if len(split) > 1 {
|
|
|
|
i.prefix = path.Join(split[:len(split)-1]...)
|
|
|
|
} else {
|
|
|
|
i.prefix = ""
|
|
|
|
}
|
|
|
|
// Object name is last element
|
|
|
|
i.objectName = split[len(split)-1]
|
|
|
|
}
|
|
|
|
|
|
|
|
// actionMeta contains information used to apply actions.
|
|
|
|
type actionMeta struct {
|
2020-09-10 03:11:24 +02:00
|
|
|
oi ObjectInfo
|
|
|
|
successorModTime time.Time // The modtime of the successor version
|
|
|
|
numVersions int // The number of versions of this object
|
2020-06-12 19:28:21 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
// applyActions will apply lifecycle checks on to a scanned item.
|
|
|
|
// The resulting size on disk will always be returned.
|
|
|
|
// The metadata will be compared to consensus on the object layer before any changes are applied.
|
|
|
|
// If no metadata is supplied, -1 is returned if no action is taken.
|
|
|
|
func (i *crawlItem) applyActions(ctx context.Context, o ObjectLayer, meta actionMeta) (size int64) {
|
|
|
|
size, err := meta.oi.GetActualSize()
|
|
|
|
if i.debug {
|
|
|
|
logger.LogIf(ctx, err)
|
|
|
|
}
|
2020-08-24 22:47:01 +02:00
|
|
|
if i.heal {
|
|
|
|
if i.debug {
|
|
|
|
logger.Info(color.Green("applyActions:")+" heal checking: %v/%v v%s", i.bucket, i.objectPath(), meta.oi.VersionID)
|
|
|
|
}
|
|
|
|
res, err := o.HealObject(ctx, i.bucket, i.objectPath(), meta.oi.VersionID, madmin.HealOpts{Remove: healDeleteDangling})
|
|
|
|
if isErrObjectNotFound(err) || isErrVersionNotFound(err) {
|
|
|
|
return 0
|
|
|
|
}
|
2020-09-12 09:08:12 +02:00
|
|
|
if err != nil && !errors.Is(err, NotImplemented{}) {
|
2020-08-24 22:47:01 +02:00
|
|
|
logger.LogIf(ctx, err)
|
|
|
|
return 0
|
|
|
|
}
|
|
|
|
size = res.ObjectSize
|
|
|
|
}
|
2020-06-12 19:28:21 +02:00
|
|
|
if i.lifeCycle == nil {
|
|
|
|
return size
|
|
|
|
}
|
|
|
|
|
2020-06-13 05:04:01 +02:00
|
|
|
versionID := meta.oi.VersionID
|
|
|
|
action := i.lifeCycle.ComputeAction(
|
|
|
|
lifecycle.ObjectOpts{
|
2020-09-10 03:11:24 +02:00
|
|
|
Name: i.objectPath(),
|
|
|
|
UserTags: meta.oi.UserTags,
|
|
|
|
ModTime: meta.oi.ModTime,
|
|
|
|
VersionID: meta.oi.VersionID,
|
|
|
|
DeleteMarker: meta.oi.DeleteMarker,
|
|
|
|
IsLatest: meta.oi.IsLatest,
|
|
|
|
NumVersions: meta.numVersions,
|
|
|
|
SuccessorModTime: meta.successorModTime,
|
2020-11-12 21:12:09 +01:00
|
|
|
RestoreOngoing: meta.oi.RestoreOngoing,
|
|
|
|
RestoreExpires: meta.oi.RestoreExpires,
|
|
|
|
TransitionStatus: meta.oi.TransitionStatus,
|
2020-06-13 05:04:01 +02:00
|
|
|
})
|
2020-06-12 19:28:21 +02:00
|
|
|
if i.debug {
|
2020-08-04 08:04:40 +02:00
|
|
|
logger.Info(color.Green("applyActions:")+" lifecycle: %q (version-id=%s), Initial scan: %v", i.objectPath(), versionID, action)
|
2020-06-12 19:28:21 +02:00
|
|
|
}
|
|
|
|
switch action {
|
2020-07-05 05:56:02 +02:00
|
|
|
case lifecycle.DeleteAction, lifecycle.DeleteVersionAction:
|
2020-11-12 21:12:09 +01:00
|
|
|
case lifecycle.TransitionAction, lifecycle.TransitionVersionAction:
|
|
|
|
case lifecycle.DeleteRestoredAction, lifecycle.DeleteRestoredVersionAction:
|
2020-06-12 19:28:21 +02:00
|
|
|
default:
|
|
|
|
// No action.
|
|
|
|
return size
|
|
|
|
}
|
|
|
|
|
2020-08-04 08:04:40 +02:00
|
|
|
obj, err := o.GetObjectInfo(ctx, i.bucket, i.objectPath(), ObjectOptions{
|
|
|
|
VersionID: versionID,
|
|
|
|
})
|
|
|
|
if err != nil {
|
|
|
|
switch err.(type) {
|
|
|
|
case MethodNotAllowed: // This happens usually for a delete marker
|
|
|
|
if !obj.DeleteMarker { // if this is not a delete marker log and return
|
|
|
|
// Do nothing - heal in the future.
|
2020-06-13 05:04:01 +02:00
|
|
|
logger.LogIf(ctx, err)
|
|
|
|
return size
|
|
|
|
}
|
2020-08-04 08:04:40 +02:00
|
|
|
case ObjectNotFound:
|
|
|
|
// object not found return 0
|
|
|
|
return 0
|
2020-06-12 19:28:21 +02:00
|
|
|
default:
|
2020-08-04 08:04:40 +02:00
|
|
|
// All other errors proceed.
|
|
|
|
logger.LogIf(ctx, err)
|
2020-06-12 19:28:21 +02:00
|
|
|
return size
|
|
|
|
}
|
|
|
|
}
|
2020-08-04 08:04:40 +02:00
|
|
|
size = obj.Size
|
|
|
|
|
|
|
|
// Recalculate action.
|
2020-11-12 21:12:09 +01:00
|
|
|
lcOpts := lifecycle.ObjectOpts{
|
|
|
|
Name: i.objectPath(),
|
|
|
|
UserTags: obj.UserTags,
|
|
|
|
ModTime: obj.ModTime,
|
|
|
|
VersionID: obj.VersionID,
|
|
|
|
DeleteMarker: obj.DeleteMarker,
|
|
|
|
IsLatest: obj.IsLatest,
|
|
|
|
NumVersions: meta.numVersions,
|
|
|
|
SuccessorModTime: meta.successorModTime,
|
|
|
|
RestoreOngoing: obj.RestoreOngoing,
|
|
|
|
RestoreExpires: obj.RestoreExpires,
|
|
|
|
TransitionStatus: obj.TransitionStatus,
|
|
|
|
}
|
|
|
|
action = i.lifeCycle.ComputeAction(lcOpts)
|
|
|
|
|
2020-08-04 08:04:40 +02:00
|
|
|
if i.debug {
|
|
|
|
logger.Info(color.Green("applyActions:")+" lifecycle: Secondary scan: %v", action)
|
|
|
|
}
|
|
|
|
switch action {
|
|
|
|
case lifecycle.DeleteAction, lifecycle.DeleteVersionAction:
|
2020-11-12 21:12:09 +01:00
|
|
|
case lifecycle.TransitionAction, lifecycle.TransitionVersionAction:
|
|
|
|
case lifecycle.DeleteRestoredAction, lifecycle.DeleteRestoredVersionAction:
|
2020-08-04 08:04:40 +02:00
|
|
|
default:
|
|
|
|
// No action.
|
|
|
|
return size
|
|
|
|
}
|
2020-06-12 19:28:21 +02:00
|
|
|
|
2020-07-05 05:56:02 +02:00
|
|
|
opts := ObjectOptions{}
|
|
|
|
switch action {
|
2020-11-12 21:12:09 +01:00
|
|
|
case lifecycle.DeleteVersionAction, lifecycle.DeleteRestoredVersionAction:
|
2020-08-04 08:04:40 +02:00
|
|
|
// Defensive code, should never happen
|
|
|
|
if obj.VersionID == "" {
|
|
|
|
return size
|
|
|
|
}
|
|
|
|
if rcfg, _ := globalBucketObjectLockSys.Get(i.bucket); rcfg.LockEnabled {
|
|
|
|
locked := enforceRetentionForDeletion(ctx, obj)
|
|
|
|
if locked {
|
|
|
|
if i.debug {
|
|
|
|
logger.Info(color.Green("applyActions:")+" lifecycle: %s is locked, not deleting", i.objectPath())
|
|
|
|
}
|
|
|
|
return size
|
|
|
|
}
|
|
|
|
}
|
|
|
|
opts.VersionID = obj.VersionID
|
2020-11-12 21:12:09 +01:00
|
|
|
case lifecycle.DeleteAction, lifecycle.DeleteRestoredAction:
|
2020-07-05 05:56:02 +02:00
|
|
|
opts.Versioned = globalBucketVersioningSys.Enabled(i.bucket)
|
2020-11-12 21:12:09 +01:00
|
|
|
case lifecycle.TransitionAction, lifecycle.TransitionVersionAction:
|
|
|
|
if obj.TransitionStatus == "" {
|
|
|
|
opts.Versioned = globalBucketVersioningSys.Enabled(obj.Bucket)
|
|
|
|
opts.VersionID = obj.VersionID
|
|
|
|
opts.TransitionStatus = lifecycle.TransitionPending
|
|
|
|
if _, err = o.DeleteObject(ctx, obj.Bucket, obj.Name, opts); err != nil {
|
|
|
|
// Assume it is still there.
|
|
|
|
logger.LogIf(ctx, err)
|
|
|
|
return size
|
|
|
|
}
|
|
|
|
}
|
|
|
|
globalTransitionState.queueTransitionTask(obj)
|
|
|
|
return 0
|
2020-07-05 05:56:02 +02:00
|
|
|
}
|
2020-11-12 21:12:09 +01:00
|
|
|
if obj.TransitionStatus != "" {
|
2020-11-25 20:24:50 +01:00
|
|
|
if err := deleteTransitionedObject(ctx, o, i.bucket, i.objectPath(), lcOpts, action, false); err != nil {
|
2020-11-12 21:12:09 +01:00
|
|
|
logger.LogIf(ctx, err)
|
|
|
|
return size
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
obj, err = o.DeleteObject(ctx, i.bucket, i.objectPath(), opts)
|
|
|
|
if err != nil {
|
|
|
|
// Assume it is still there.
|
|
|
|
logger.LogIf(ctx, err)
|
|
|
|
return size
|
|
|
|
}
|
2020-06-12 19:28:21 +02:00
|
|
|
}
|
|
|
|
|
2020-10-17 06:22:12 +02:00
|
|
|
eventName := event.ObjectRemovedDelete
|
|
|
|
if obj.DeleteMarker {
|
|
|
|
eventName = event.ObjectRemovedDeleteMarkerCreated
|
|
|
|
}
|
|
|
|
|
2020-06-12 19:28:21 +02:00
|
|
|
// Notify object deleted event.
|
|
|
|
sendEvent(eventArgs{
|
2020-10-17 06:22:12 +02:00
|
|
|
EventName: eventName,
|
2020-06-12 19:28:21 +02:00
|
|
|
BucketName: i.bucket,
|
2020-06-13 05:04:01 +02:00
|
|
|
Object: obj,
|
|
|
|
Host: "Internal: [ILM-EXPIRY]",
|
2020-06-12 19:28:21 +02:00
|
|
|
})
|
|
|
|
return 0
|
|
|
|
}
|
|
|
|
|
|
|
|
// objectPath returns the prefix and object name.
|
|
|
|
func (i *crawlItem) objectPath() string {
|
|
|
|
return path.Join(i.prefix, i.objectName)
|
|
|
|
}
|
|
|
|
|
2020-11-12 18:17:56 +01:00
|
|
|
// sleepDuration multiplies the duration d by x
|
|
|
|
// and sleeps if is more than 100 micro seconds.
|
|
|
|
// Sleep is limited to max 15 seconds.
|
2020-06-12 19:28:21 +02:00
|
|
|
func sleepDuration(d time.Duration, x float64) {
|
2020-11-12 18:17:56 +01:00
|
|
|
const maxWait = 15 * time.Second
|
|
|
|
const minWait = 100 * time.Microsecond
|
2020-06-12 19:28:21 +02:00
|
|
|
// Don't sleep for really small amount of time
|
2020-11-12 18:17:56 +01:00
|
|
|
if d := time.Duration(float64(d) * x); d > minWait {
|
|
|
|
if d > maxWait {
|
|
|
|
d = maxWait
|
2020-06-12 19:28:21 +02:00
|
|
|
}
|
|
|
|
time.Sleep(d)
|
|
|
|
}
|
|
|
|
}
|
2020-07-22 02:49:56 +02:00
|
|
|
|
|
|
|
// healReplication will heal a scanned item that has failed replication.
|
|
|
|
func (i *crawlItem) healReplication(ctx context.Context, o ObjectLayer, meta actionMeta) {
|
2020-11-20 03:43:58 +01:00
|
|
|
if meta.oi.DeleteMarker || !meta.oi.VersionPurgeStatus.Empty() {
|
|
|
|
//heal delete marker replication failure or versioned delete replication failure
|
|
|
|
if meta.oi.ReplicationStatus == replication.Pending ||
|
|
|
|
meta.oi.ReplicationStatus == replication.Failed ||
|
|
|
|
meta.oi.VersionPurgeStatus == Failed || meta.oi.VersionPurgeStatus == Pending {
|
|
|
|
i.healReplicationDeletes(ctx, o, meta)
|
|
|
|
return
|
|
|
|
}
|
|
|
|
}
|
2020-07-22 02:49:56 +02:00
|
|
|
if meta.oi.ReplicationStatus == replication.Pending ||
|
|
|
|
meta.oi.ReplicationStatus == replication.Failed {
|
2020-09-17 01:04:55 +02:00
|
|
|
globalReplicationState.queueReplicaTask(meta.oi)
|
2020-07-22 02:49:56 +02:00
|
|
|
}
|
|
|
|
}
|
2020-11-20 03:43:58 +01:00
|
|
|
|
|
|
|
// healReplicationDeletes will heal a scanned deleted item that failed to replicate deletes.
|
|
|
|
func (i *crawlItem) healReplicationDeletes(ctx context.Context, o ObjectLayer, meta actionMeta) {
|
|
|
|
// handle soft delete and permanent delete failures here.
|
|
|
|
if meta.oi.DeleteMarker || !meta.oi.VersionPurgeStatus.Empty() {
|
|
|
|
versionID := ""
|
|
|
|
dmVersionID := ""
|
|
|
|
if meta.oi.VersionPurgeStatus.Empty() {
|
|
|
|
dmVersionID = meta.oi.VersionID
|
|
|
|
} else {
|
|
|
|
versionID = meta.oi.VersionID
|
|
|
|
}
|
|
|
|
globalReplicationState.queueReplicaDeleteTask(DeletedObjectVersionInfo{
|
|
|
|
DeletedObject: DeletedObject{
|
|
|
|
ObjectName: meta.oi.Name,
|
|
|
|
DeleteMarkerVersionID: dmVersionID,
|
|
|
|
VersionID: versionID,
|
|
|
|
DeleteMarkerReplicationStatus: string(meta.oi.ReplicationStatus),
|
|
|
|
DeleteMarkerMTime: meta.oi.ModTime,
|
|
|
|
DeleteMarker: meta.oi.DeleteMarker,
|
|
|
|
VersionPurgeStatus: meta.oi.VersionPurgeStatus,
|
|
|
|
},
|
|
|
|
Bucket: meta.oi.Bucket,
|
|
|
|
})
|
|
|
|
}
|
|
|
|
}
|