minio/cmd/disk-cache-fs.go
poornas a3e806ed61 Add disk based edge caching support. (#5182)
This PR adds disk based edge caching support for minio server.

Cache settings can be configured in config.json to take list of disk drives,
cache expiry in days and file patterns to exclude from cache or via environment
variables MINIO_CACHE_DRIVES, MINIO_CACHE_EXCLUDE and MINIO_CACHE_EXPIRY

Design assumes that Atime support is enabled and the list of cache drives is
fixed.
 - Objects are cached on both GET and PUT/POST operations.
 - Expiry is used as hint to evict older entries from cache, or if 80% of cache
   capacity is filled.
 - When object storage backend is down, GET, LIST and HEAD operations fetch
   object seamlessly from cache.

Current Limitations
 - Bucket policies are not cached, so anonymous operations are not supported in
   offline mode.
 - Objects are distributed using deterministic hashing among list of cache
   drives specified.If one or more drives go offline, or cache drive
   configuration is altered - performance could degrade to linear lookup.

Fixes #4026
2018-03-28 14:14:06 -07:00

507 lines
15 KiB
Go

/*
* Minio Cloud Storage, (C) 2018 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 (
"context"
"encoding/hex"
"encoding/json"
"io"
"io/ioutil"
"os"
"path"
"sync"
"time"
"github.com/minio/minio/pkg/disk"
errors2 "github.com/minio/minio/pkg/errors"
"github.com/minio/minio/pkg/hash"
"github.com/minio/minio/pkg/lock"
)
const (
// cache.json object metadata for cached objects.
cacheMetaJSONFile = "cache.json"
cacheMetaFormat = "cache"
)
// cacheFSObjects implements the cache backend operations.
type cacheFSObjects struct {
*FSObjects
// caching drive path (from cache "drives" in config.json)
dir string
// expiry in days specified in config.json
expiry int
// max disk usage pct
maxDiskUsagePct int
// purge() listens on this channel to start the cache-purge process
purgeChan chan struct{}
// mark false if drive is offline
online bool
// mutex to protect updates to online variable
onlineMutex *sync.RWMutex
}
// Inits the cache directory if it is not init'ed already.
// Initializing implies creation of new FS Object layer.
func newCacheFSObjects(dir string, expiry int, maxDiskUsagePct int) (*cacheFSObjects, error) {
obj, err := newFSObjects(dir, cacheMetaJSONFile)
if err != nil {
return nil, err
}
trashPath := pathJoin(dir, minioMetaBucket, cacheTrashDir)
if err := os.MkdirAll(trashPath, 0777); err != nil {
return nil, err
}
if expiry == 0 {
expiry = globalCacheExpiry
}
var cacheFS cacheFSObjects
fsObjects := obj.(*FSObjects)
cacheFS = cacheFSObjects{
FSObjects: fsObjects,
dir: dir,
expiry: expiry,
maxDiskUsagePct: maxDiskUsagePct,
purgeChan: make(chan struct{}),
online: true,
onlineMutex: &sync.RWMutex{},
}
return &cacheFS, nil
}
// Returns if the disk usage is low.
// Disk usage is low if usage is < 80% of cacheMaxDiskUsagePct
// Ex. for a 100GB disk, if maxUsage is configured as 70% then cacheMaxDiskUsagePct is 70G
// hence disk usage is low if the disk usage is less than 56G (because 80% of 70G is 56G)
func (cfs *cacheFSObjects) diskUsageLow() bool {
minUsage := cfs.maxDiskUsagePct * 80 / 100
di, err := disk.GetInfo(cfs.dir)
if err != nil {
errorIf(err, "Error getting disk information on %s", cfs.dir)
return false
}
usedPercent := (di.Total - di.Free) * 100 / di.Total
return int(usedPercent) < minUsage
}
// Return if the disk usage is high.
// Disk usage is high if disk used is > cacheMaxDiskUsagePct
func (cfs *cacheFSObjects) diskUsageHigh() bool {
di, err := disk.GetInfo(cfs.dir)
if err != nil {
errorIf(err, "Error getting disk information on %s", cfs.dir)
return true
}
usedPercent := (di.Total - di.Free) * 100 / di.Total
return int(usedPercent) > cfs.maxDiskUsagePct
}
// Returns if size space can be allocated without exceeding
// max disk usable for caching
func (cfs *cacheFSObjects) diskAvailable(size int64) bool {
di, err := disk.GetInfo(cfs.dir)
if err != nil {
errorIf(err, "Error getting disk information on %s", cfs.dir)
return false
}
usedPercent := (di.Total - (di.Free - uint64(size))) * 100 / di.Total
return int(usedPercent) < cfs.maxDiskUsagePct
}
// purges all content marked trash from the cache.
func (cfs *cacheFSObjects) purgeTrash() {
ticker := time.NewTicker(time.Minute * cacheCleanupInterval)
for {
select {
case <-globalServiceDoneCh:
// Stop the timer.
ticker.Stop()
return
case <-ticker.C:
trashPath := path.Join(cfs.fsPath, minioMetaBucket, cacheTrashDir)
entries, err := readDir(trashPath)
if err != nil {
return
}
for _, entry := range entries {
fi, err := fsStatVolume(pathJoin(trashPath, entry))
if err != nil {
continue
}
dir := path.Join(trashPath, fi.Name())
// Delete all expired cache content.
fsRemoveAll(dir)
}
}
}
}
// Purge cache entries that were not accessed.
func (cfs *cacheFSObjects) purge() {
delimiter := slashSeparator
maxKeys := 1000
ctx := context.Background()
for {
olderThan := cfs.expiry
for !cfs.diskUsageLow() {
// delete unaccessed objects older than expiry duration
expiry := UTCNow().AddDate(0, 0, -1*olderThan)
olderThan /= 2
if olderThan < 1 {
break
}
deletedCount := 0
buckets, err := cfs.ListBuckets(ctx)
if err != nil {
errorIf(err, "Unable to list buckets.")
}
// Reset cache online status if drive was offline earlier.
if !cfs.IsOnline() {
cfs.setOnline(true)
}
for _, bucket := range buckets {
var continuationToken string
var marker string
for {
objects, err := cfs.ListObjects(ctx, bucket.Name, marker, continuationToken, delimiter, maxKeys)
if err != nil {
break
}
if !objects.IsTruncated {
break
}
marker = objects.NextMarker
for _, object := range objects.Objects {
// purge objects that qualify because of cache-control directives or
// past cache expiry duration.
if !filterFromCache(object.UserDefined) ||
!isStaleCache(object) ||
object.AccTime.After(expiry) {
continue
}
if err = cfs.DeleteObject(ctx, bucket.Name, object.Name); err != nil {
errorIf(err, "Unable to remove cache entry in dir %s/%s", bucket.Name, object.Name)
continue
}
deletedCount++
}
}
}
if deletedCount == 0 {
// to avoid a busy loop
time.Sleep(time.Minute * 30)
}
}
<-cfs.purgeChan
}
}
// sets cache drive status
func (cfs *cacheFSObjects) setOnline(status bool) {
cfs.onlineMutex.Lock()
cfs.online = status
cfs.onlineMutex.Unlock()
}
// returns true if cache drive is online
func (cfs *cacheFSObjects) IsOnline() bool {
cfs.onlineMutex.RLock()
defer cfs.onlineMutex.RUnlock()
return cfs.online
}
// Caches the object to disk
func (cfs *cacheFSObjects) Put(ctx context.Context, bucket, object string, data *hash.Reader, metadata map[string]string) error {
if cfs.diskUsageHigh() {
select {
case cfs.purgeChan <- struct{}{}:
default:
}
return errDiskFull
}
if !cfs.diskAvailable(data.Size()) {
return errDiskFull
}
if _, err := cfs.GetBucketInfo(ctx, bucket); err != nil {
pErr := cfs.MakeBucketWithLocation(ctx, bucket, "")
if pErr != nil {
return pErr
}
}
_, err := cfs.PutObject(ctx, bucket, object, data, metadata)
// if err is due to disk being offline , mark cache drive as offline
if errors2.IsErr(err, baseErrs...) {
cfs.setOnline(false)
}
return err
}
// Returns the handle for the cached object
func (cfs *cacheFSObjects) Get(ctx context.Context, bucket, object string, startOffset int64, length int64, writer io.Writer, etag string) (err error) {
return cfs.GetObject(ctx, bucket, object, startOffset, length, writer, etag)
}
// Deletes the cached object
func (cfs *cacheFSObjects) Delete(ctx context.Context, bucket, object string) (err error) {
return cfs.DeleteObject(ctx, bucket, object)
}
// convenience function to check if object is cached on this cacheFSObjects
func (cfs *cacheFSObjects) Exists(ctx context.Context, bucket, object string) bool {
_, err := cfs.GetObjectInfo(ctx, bucket, object)
return err == nil
}
// Identical to fs PutObject operation except that it uses ETag in metadata
// headers.
func (cfs *cacheFSObjects) PutObject(ctx context.Context, bucket string, object string, data *hash.Reader, metadata map[string]string) (objInfo ObjectInfo, retErr error) {
fs := cfs.FSObjects
// Lock the object.
objectLock := fs.nsMutex.NewNSLock(bucket, object)
if err := objectLock.GetLock(globalObjectTimeout); err != nil {
return objInfo, err
}
defer objectLock.Unlock()
// No metadata is set, allocate a new one.
meta := make(map[string]string)
for k, v := range metadata {
meta[k] = v
}
var err error
// Validate if bucket name is valid and exists.
if _, err = fs.statBucketDir(bucket); err != nil {
return ObjectInfo{}, toObjectErr(err, bucket)
}
fsMeta := newFSMetaV1()
fsMeta.Meta = meta
// This is a special case with size as '0' and object ends
// with a slash separator, we treat it like a valid operation
// and return success.
if isObjectDir(object, data.Size()) {
// Check if an object is present as one of the parent dir.
if fs.parentDirIsObject(bucket, path.Dir(object)) {
return ObjectInfo{}, toObjectErr(errors2.Trace(errFileAccessDenied), bucket, object)
}
if err = mkdirAll(pathJoin(fs.fsPath, bucket, object), 0777); err != nil {
return ObjectInfo{}, toObjectErr(err, bucket, object)
}
var fi os.FileInfo
if fi, err = fsStatDir(pathJoin(fs.fsPath, bucket, object)); err != nil {
return ObjectInfo{}, toObjectErr(err, bucket, object)
}
return fsMeta.ToObjectInfo(bucket, object, fi), nil
}
if err = checkPutObjectArgs(bucket, object, fs, data.Size()); err != nil {
return ObjectInfo{}, err
}
// Check if an object is present as one of the parent dir.
if fs.parentDirIsObject(bucket, path.Dir(object)) {
return ObjectInfo{}, toObjectErr(errors2.Trace(errFileAccessDenied), bucket, object)
}
// Validate input data size and it can never be less than zero.
if data.Size() < 0 {
return ObjectInfo{}, errors2.Trace(errInvalidArgument)
}
var wlk *lock.LockedFile
if bucket != minioMetaBucket {
bucketMetaDir := pathJoin(fs.fsPath, minioMetaBucket, bucketMetaPrefix)
fsMetaPath := pathJoin(bucketMetaDir, bucket, object, fs.metaJSONFile)
wlk, err = fs.rwPool.Create(fsMetaPath)
if err != nil {
return ObjectInfo{}, toObjectErr(errors2.Trace(err), bucket, object)
}
// This close will allow for locks to be synchronized on `fs.json`.
defer wlk.Close()
defer func() {
// Remove meta file when PutObject encounters any error
if retErr != nil {
tmpDir := pathJoin(fs.fsPath, minioMetaTmpBucket, fs.fsUUID)
fsRemoveMeta(bucketMetaDir, fsMetaPath, tmpDir)
}
}()
}
// Uploaded object will first be written to the temporary location which will eventually
// be renamed to the actual location. It is first written to the temporary location
// so that cleaning it up will be easy if the server goes down.
tempObj := mustGetUUID()
// Allocate a buffer to Read() from request body
bufSize := int64(readSizeV1)
if size := data.Size(); size > 0 && bufSize > size {
bufSize = size
}
buf := make([]byte, int(bufSize))
fsTmpObjPath := pathJoin(fs.fsPath, minioMetaTmpBucket, fs.fsUUID, tempObj)
bytesWritten, err := fsCreateFile(fsTmpObjPath, data, buf, data.Size())
if err != nil {
fsRemoveFile(fsTmpObjPath)
errorIf(err, "Failed to create object %s/%s", bucket, object)
return ObjectInfo{}, toObjectErr(err, bucket, object)
}
if fsMeta.Meta["etag"] == "" {
fsMeta.Meta["etag"] = hex.EncodeToString(data.MD5Current())
}
// Should return IncompleteBody{} error when reader has fewer
// bytes than specified in request header.
if bytesWritten < data.Size() {
fsRemoveFile(fsTmpObjPath)
return ObjectInfo{}, errors2.Trace(IncompleteBody{})
}
// Delete the temporary object in the case of a
// failure. If PutObject succeeds, then there would be
// nothing to delete.
defer fsRemoveFile(fsTmpObjPath)
// Entire object was written to the temp location, now it's safe to rename it to the actual location.
fsNSObjPath := pathJoin(fs.fsPath, bucket, object)
if err = fsRenameFile(fsTmpObjPath, fsNSObjPath); err != nil {
return ObjectInfo{}, toObjectErr(err, bucket, object)
}
if bucket != minioMetaBucket {
// Write FS metadata after a successful namespace operation.
if _, err = fsMeta.WriteTo(wlk); err != nil {
return ObjectInfo{}, toObjectErr(err, bucket, object)
}
}
// Stat the file to fetch timestamp, size.
fi, err := fsStatFile(pathJoin(fs.fsPath, bucket, object))
if err != nil {
return ObjectInfo{}, toObjectErr(err, bucket, object)
}
// Success.
return fsMeta.ToObjectInfo(bucket, object, fi), nil
}
// Implements S3 compatible initiate multipart API. Operation here is identical
// to fs backend implementation - with the exception that cache FS uses the uploadID
// generated on the backend
func (cfs *cacheFSObjects) NewMultipartUpload(ctx context.Context, bucket, object string, meta map[string]string, uploadID string) (string, error) {
if cfs.diskUsageHigh() {
select {
case cfs.purgeChan <- struct{}{}:
default:
}
return "", errDiskFull
}
if _, err := cfs.GetBucketInfo(ctx, bucket); err != nil {
pErr := cfs.MakeBucketWithLocation(ctx, bucket, "")
if pErr != nil {
return "", pErr
}
}
fs := cfs.FSObjects
if err := checkNewMultipartArgs(bucket, object, fs); err != nil {
return "", toObjectErr(err, bucket)
}
if _, err := fs.statBucketDir(bucket); err != nil {
return "", toObjectErr(err, bucket)
}
uploadIDDir := fs.getUploadIDDir(bucket, object, uploadID)
err := mkdirAll(uploadIDDir, 0755)
if err != nil {
return "", errors2.Trace(err)
}
// Initialize fs.json values.
fsMeta := newFSMetaV1()
fsMeta.Meta = meta
fsMetaBytes, err := json.Marshal(fsMeta)
if err != nil {
return "", errors2.Trace(err)
}
if err = ioutil.WriteFile(pathJoin(uploadIDDir, fs.metaJSONFile), fsMetaBytes, 0644); err != nil {
return "", errors2.Trace(err)
}
return uploadID, nil
}
// moveBucketToTrash clears cacheFSObjects of bucket contents and moves it to trash folder.
func (cfs *cacheFSObjects) moveBucketToTrash(ctx context.Context, bucket string) (err error) {
fs := cfs.FSObjects
bucketLock := fs.nsMutex.NewNSLock(bucket, "")
if err = bucketLock.GetLock(globalObjectTimeout); err != nil {
return err
}
defer bucketLock.Unlock()
bucketDir, err := fs.getBucketDir(bucket)
if err != nil {
return toObjectErr(err, bucket)
}
trashPath := pathJoin(cfs.fsPath, minioMetaBucket, cacheTrashDir)
expiredDir := path.Join(trashPath, bucket)
// Attempt to move regular bucket to expired directory.
if err = fsRenameDir(bucketDir, expiredDir); err != nil {
return toObjectErr(err, bucket)
}
// Cleanup all the bucket metadata.
ominioMetadataBucketDir := pathJoin(fs.fsPath, minioMetaBucket, bucketMetaPrefix, bucket)
nminioMetadataBucketDir := pathJoin(trashPath, MustGetUUID())
_ = fsRenameDir(ominioMetadataBucketDir, nminioMetadataBucketDir)
return nil
}
// Removes a directory only if its empty, handles long
// paths for windows automatically.
func fsRenameDir(dirPath, newPath string) (err error) {
if dirPath == "" || newPath == "" {
return errors2.Trace(errInvalidArgument)
}
if err = checkPathLength(dirPath); err != nil {
return errors2.Trace(err)
}
if err = checkPathLength(newPath); err != nil {
return errors2.Trace(err)
}
if err = os.Rename(dirPath, newPath); err != nil {
if os.IsNotExist(err) {
return errors2.Trace(errVolumeNotFound)
} else if isSysErrNotEmpty(err) {
return errors2.Trace(errVolumeNotEmpty)
}
return errors2.Trace(err)
}
return nil
}