minio/cmd/data-usage-cache.go
Klaus Post 073aac3d92
add data update tracking using bloom filter (#9208)
By monitoring PUT/DELETE and heal operations it is possible
to track changed paths and keep a bloom filter for this data. 

This can help prioritize paths to scan. The bloom filter can identify
paths that have not changed, and the few collisions will only result
in a marginal extra workload. This can be implemented on either a
bucket+(1 prefix level) with reasonable performance.

The bloom filter is set to have a false positive rate at 1% at 1M 
entries. A bloom table of this size is about ~2500 bytes when serialized.

To not force a full scan of all paths that have changed cycle bloom
filters would need to be kept, so we guarantee that dirty paths have
been scanned within cycle runs. Until cycle bloom filters have been
collected all paths are considered dirty.
2020-04-27 10:06:21 -07:00

553 lines
14 KiB
Go

/*
* 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"
"fmt"
"io"
"path"
"path/filepath"
"strings"
"time"
"github.com/cespare/xxhash/v2"
"github.com/minio/minio/cmd/logger"
"github.com/minio/minio/pkg/hash"
"github.com/tinylib/msgp/msgp"
)
const dataUsageHashLen = 8
//go:generate msgp -file $GOFILE -unexported
// dataUsageHash is the hash type used.
type dataUsageHash uint64
// sizeHistogram is a size histogram.
type sizeHistogram [dataUsageBucketLen]uint64
//msgp:tuple dataUsageEntry
type dataUsageEntry struct {
// These fields do no include any children.
Size int64
Objects uint64
ObjSizes sizeHistogram
Children dataUsageHashMap
}
//msgp:ignore dataUsageEntryInfo
type dataUsageEntryInfo struct {
Name string
Parent string
Entry dataUsageEntry
}
type dataUsageCacheInfo struct {
// Name of the bucket. Also root element.
Name string
LastUpdate time.Time
NextCycle uint32
BloomFilter []byte `msg:"BloomFilter,omitempty"`
}
// merge other data usage entry into this, excluding children.
func (e *dataUsageEntry) merge(other dataUsageEntry) {
e.Objects += other.Objects
e.Size += other.Size
for i, v := range other.ObjSizes[:] {
e.ObjSizes[i] += v
}
}
// mod returns true if the hash mod cycles == cycle.
func (h dataUsageHash) mod(cycle uint32, cycles uint32) bool {
return uint32(h)%cycles == cycle%cycles
}
// addChildString will add a child based on its name.
// If it already exists it will not be added again.
func (e *dataUsageEntry) addChildString(name string) {
e.addChild(hashPath(name))
}
// addChild will add a child based on its hash.
// If it already exists it will not be added again.
func (e *dataUsageEntry) addChild(hash dataUsageHash) {
if _, ok := e.Children[hash]; ok {
return
}
if e.Children == nil {
e.Children = make(dataUsageHashMap, 1)
}
e.Children[hash] = struct{}{}
}
// find a path in the cache.
// Returns nil if not found.
func (d *dataUsageCache) find(path string) *dataUsageEntry {
due, ok := d.Cache[hashPath(path)]
if !ok {
return nil
}
return &due
}
// dui converts the flattened version of the path to DataUsageInfo.
// As a side effect d will be flattened, use a clone if this is not ok.
func (d *dataUsageCache) dui(path string, buckets []BucketInfo) DataUsageInfo {
e := d.find(path)
if e == nil {
return DataUsageInfo{LastUpdate: UTCNow()}
}
flat := d.flatten(*e)
return DataUsageInfo{
LastUpdate: d.Info.LastUpdate,
ObjectsCount: flat.Objects,
ObjectsTotalSize: uint64(flat.Size),
ObjectsSizesHistogram: flat.ObjSizes.asMap(),
BucketsCount: uint64(len(e.Children)),
BucketsSizes: d.pathSizes(buckets),
}
}
// replace will add or replace an entry in the cache.
// If a parent is specified it will be added to that if not already there.
// If the parent does not exist, it will be added.
func (d *dataUsageCache) replace(path, parent string, e dataUsageEntry) {
hash := hashPath(path)
if d.Cache == nil {
d.Cache = make(map[dataUsageHash]dataUsageEntry, 100)
}
d.Cache[hash] = e
if parent != "" {
phash := hashPath(parent)
p := d.Cache[phash]
p.addChild(hash)
d.Cache[phash] = p
}
}
// replaceHashed add or replaces an entry to the cache based on its hash.
// If a parent is specified it will be added to that if not already there.
// If the parent does not exist, it will be added.
func (d *dataUsageCache) replaceHashed(hash dataUsageHash, parent *dataUsageHash, e dataUsageEntry) {
if d.Cache == nil {
d.Cache = make(map[dataUsageHash]dataUsageEntry, 100)
}
d.Cache[hash] = e
if parent != nil {
p := d.Cache[*parent]
p.addChild(hash)
d.Cache[*parent] = p
}
}
// copyWithChildren will copy entry with hash from src if it exists along with any children.
// If a parent is specified it will be added to that if not already there.
// If the parent does not exist, it will be added.
func (d *dataUsageCache) copyWithChildren(src *dataUsageCache, hash dataUsageHash, parent *dataUsageHash) {
if d.Cache == nil {
d.Cache = make(map[dataUsageHash]dataUsageEntry, 100)
}
e, ok := src.Cache[hash]
if !ok {
return
}
d.Cache[hash] = e
for ch := range e.Children {
if ch == hash {
logger.LogIf(GlobalContext, errors.New("dataUsageCache.copyWithChildren: Circular reference"))
return
}
d.copyWithChildren(src, ch, &hash)
}
if parent != nil {
p := d.Cache[*parent]
p.addChild(hash)
d.Cache[*parent] = p
}
}
// StringAll returns a detailed string representation of all entries in the cache.
func (d *dataUsageCache) StringAll() string {
s := fmt.Sprintf("info:%+v\n", d.Info)
for k, v := range d.Cache {
s += fmt.Sprintf("\t%v: %+v\n", k, v)
}
return strings.TrimSpace(s)
}
// insert the hash into dst.
// dst must be at least dataUsageHashLen bytes long.
func (h dataUsageHash) bytes(dst []byte) {
binary.LittleEndian.PutUint64(dst, uint64(h))
}
// String returns a human readable representation of the string.
func (h dataUsageHash) String() string {
return fmt.Sprintf("%x", uint64(h))
}
// flatten all children of the root into the root element and return it.
func (d *dataUsageCache) flatten(root dataUsageEntry) dataUsageEntry {
for id := range root.Children {
e := d.Cache[id]
if len(e.Children) > 0 {
e = d.flatten(e)
}
root.merge(e)
}
root.Children = nil
return root
}
// add a size to the histogram.
func (h *sizeHistogram) add(size int64) {
// Fetch the histogram interval corresponding
// to the passed object size.
for i, interval := range ObjectsHistogramIntervals {
if size >= interval.start && size <= interval.end {
h[i]++
break
}
}
}
// asMap returns the map as a map[string]uint64.
func (h *sizeHistogram) asMap() map[string]uint64 {
res := make(map[string]uint64, 7)
for i, count := range h {
res[ObjectsHistogramIntervals[i].name] = count
}
return res
}
// pathSizes returns the path sizes as a map.
func (d *dataUsageCache) pathSizes(buckets []BucketInfo) map[string]uint64 {
var dst = make(map[string]uint64, len(buckets))
for _, bucket := range buckets {
e := d.find(bucket.Name)
if e == nil {
continue
}
flat := d.flatten(*e)
dst[bucket.Name] = uint64(flat.Size)
}
return dst
}
// sizeRecursive returns the path as a flattened entry.
func (d *dataUsageCache) sizeRecursive(path string) *dataUsageEntry {
root := d.find(path)
if root == nil || len(root.Children) == 0 {
return root
}
flat := d.flatten(*root)
return &flat
}
// dataUsageCache contains a cache of data usage entries.
//msgp:ignore dataUsageCache
type dataUsageCache struct {
Info dataUsageCacheInfo
Cache map[dataUsageHash]dataUsageEntry
}
// root returns the root of the cache.
func (d *dataUsageCache) root() *dataUsageEntry {
return d.find(d.Info.Name)
}
// rootHash returns the root of the cache.
func (d *dataUsageCache) rootHash() dataUsageHash {
return hashPath(d.Info.Name)
}
// clone returns a copy of the cache with no references to the existing.
func (d *dataUsageCache) clone() dataUsageCache {
clone := dataUsageCache{
Info: d.Info,
Cache: make(map[dataUsageHash]dataUsageEntry, len(d.Cache)),
}
for k, v := range d.Cache {
clone.Cache[k] = v
}
return clone
}
// merge root of other into d.
// children of root will be flattened before being merged.
// Last update time will be set to the last updated.
func (d *dataUsageCache) merge(other dataUsageCache) {
existingRoot := d.root()
otherRoot := other.root()
if existingRoot == nil && otherRoot == nil {
return
}
if otherRoot == nil {
return
}
if existingRoot == nil {
*d = other.clone()
return
}
if other.Info.LastUpdate.After(d.Info.LastUpdate) {
d.Info.LastUpdate = other.Info.LastUpdate
}
existingRoot.merge(*otherRoot)
eHash := d.rootHash()
for key := range otherRoot.Children {
entry := other.Cache[key]
flat := other.flatten(entry)
existing := d.Cache[key]
// If not found, merging simply adds.
existing.merge(flat)
d.replaceHashed(key, &eHash, existing)
}
}
// load the cache content with name from minioMetaBackgroundOpsBucket.
// Only backend errors are returned as errors.
// If the object is not found or unable to deserialize d is cleared and nil error is returned.
func (d *dataUsageCache) load(ctx context.Context, store ObjectLayer, name string) error {
var buf bytes.Buffer
err := store.GetObject(ctx, dataUsageBucket, name, 0, -1, &buf, "", ObjectOptions{})
if err != nil {
if !isErrObjectNotFound(err) && !isErrBucketNotFound(err) {
return toObjectErr(err, dataUsageBucket, name)
}
*d = dataUsageCache{}
return nil
}
err = d.deserialize(buf.Bytes())
if err != nil {
*d = dataUsageCache{}
logger.LogIf(ctx, err)
}
return nil
}
// save the content of the cache to minioMetaBackgroundOpsBucket with the provided name.
func (d *dataUsageCache) save(ctx context.Context, store ObjectLayer, name string) error {
b := d.serialize()
size := int64(len(b))
r, err := hash.NewReader(bytes.NewReader(b), size, "", "", size, false)
if err != nil {
return err
}
_, err = store.PutObject(ctx,
dataUsageBucket,
name,
NewPutObjReader(r, nil, nil),
ObjectOptions{})
return err
}
// dataUsageCacheVer indicates the cache version.
// Bumping the cache version will drop data from previous versions
// and write new data with the new version.
const dataUsageCacheVer = 1
// serialize the contents of the cache.
func (d *dataUsageCache) serialize() []byte {
// Alloc pessimistically
// dataUsageCacheVer
due := dataUsageEntry{}
msgLen := 1
msgLen += d.Info.Msgsize()
// len(d.Cache)
msgLen += binary.MaxVarintLen64
// Hashes (one for key, assume 1 child/node)
msgLen += len(d.Cache) * dataUsageHashLen * 2
msgLen += len(d.Cache) * due.Msgsize()
// Create destination buffer...
dst := make([]byte, 0, msgLen)
var n int
tmp := make([]byte, 1024)
// byte: version.
dst = append(dst, dataUsageCacheVer)
// Info...
dst, err := d.Info.MarshalMsg(dst)
if err != nil {
panic(err)
}
n = binary.PutUvarint(tmp, uint64(len(d.Cache)))
dst = append(dst, tmp[:n]...)
for k, v := range d.Cache {
// Put key
binary.LittleEndian.PutUint64(tmp[:dataUsageHashLen], uint64(k))
dst = append(dst, tmp[:8]...)
tmp, err = v.MarshalMsg(tmp[:0])
if err != nil {
panic(err)
}
// key, value pairs.
dst = append(dst, tmp...)
}
return dst
}
// deserialize the supplied byte slice into the cache.
func (d *dataUsageCache) deserialize(b []byte) error {
if len(b) < 1 {
return io.ErrUnexpectedEOF
}
switch b[0] {
case 1:
default:
return fmt.Errorf("dataUsageCache: unknown version: %d", int(b[0]))
}
b = b[1:]
// Info...
b, err := d.Info.UnmarshalMsg(b)
if err != nil {
return err
}
cacheLen, n := binary.Uvarint(b)
if n <= 0 {
return fmt.Errorf("dataUsageCache: reading cachelen, n <= 0 ")
}
b = b[n:]
d.Cache = make(map[dataUsageHash]dataUsageEntry, cacheLen)
for i := 0; i < int(cacheLen); i++ {
if len(b) <= dataUsageHashLen {
return io.ErrUnexpectedEOF
}
k := binary.LittleEndian.Uint64(b[:dataUsageHashLen])
b = b[dataUsageHashLen:]
var v dataUsageEntry
b, err = v.UnmarshalMsg(b)
if err != nil {
return err
}
d.Cache[dataUsageHash(k)] = v
}
return nil
}
// Trim this from start+end of hashes.
var hashPathCutSet = dataUsageRoot
func init() {
if dataUsageRoot != string(filepath.Separator) {
hashPathCutSet = dataUsageRoot + string(filepath.Separator)
}
}
// hashPath calculates a hash of the provided string.
func hashPath(data string) dataUsageHash {
if data != dataUsageRoot {
data = strings.Trim(data, hashPathCutSet)
}
data = path.Clean(data)
return dataUsageHash(xxhash.Sum64String(data))
}
//msgp:ignore dataUsageEntryInfo
type dataUsageHashMap map[dataUsageHash]struct{}
// MarshalMsg implements msgp.Marshaler
func (d dataUsageHashMap) MarshalMsg(b []byte) (o []byte, err error) {
o = msgp.Require(b, d.Msgsize())
// Write bin header manually
const mbin32 uint8 = 0xc6
sz := uint32(len(d)) * dataUsageHashLen
o = append(o, mbin32, byte(sz>>24), byte(sz>>16), byte(sz>>8), byte(sz))
var tmp [dataUsageHashLen]byte
for k := range d {
binary.LittleEndian.PutUint64(tmp[:], uint64(k))
o = append(o, tmp[:]...)
}
return
}
// Msgsize returns an upper bound estimate of the number of bytes occupied by the serialized message
func (d dataUsageHashMap) Msgsize() (s int) {
s = 5 + len(d)*dataUsageHashLen
return
}
// UnmarshalMsg implements msgp.Unmarshaler
func (d *dataUsageHashMap) UnmarshalMsg(bts []byte) (o []byte, err error) {
var hashes []byte
hashes, bts, err = msgp.ReadBytesZC(bts)
if err != nil {
err = msgp.WrapError(err, "dataUsageHashMap")
return
}
var dst = make(dataUsageHashMap, len(hashes)/dataUsageHashLen)
for len(hashes) >= dataUsageHashLen {
dst[dataUsageHash(binary.LittleEndian.Uint64(hashes[:dataUsageHashLen]))] = struct{}{}
hashes = hashes[dataUsageHashLen:]
}
*d = dst
o = bts
return
}
func (d *dataUsageHashMap) DecodeMsg(dc *msgp.Reader) (err error) {
var zb0001 uint32
zb0001, err = dc.ReadBytesHeader()
if err != nil {
err = msgp.WrapError(err)
return
}
var dst = make(dataUsageHashMap, zb0001)
var tmp [8]byte
for i := uint32(0); i < zb0001; i++ {
_, err = io.ReadFull(dc, tmp[:])
if err != nil {
err = msgp.WrapError(err, "dataUsageHashMap")
return
}
dst[dataUsageHash(binary.LittleEndian.Uint64(tmp[:]))] = struct{}{}
}
return nil
}
func (d dataUsageHashMap) EncodeMsg(en *msgp.Writer) (err error) {
err = en.WriteBytesHeader(uint32(len(d)) * dataUsageHashLen)
if err != nil {
err = msgp.WrapError(err)
return
}
var tmp [dataUsageHashLen]byte
for k := range d {
binary.LittleEndian.PutUint64(tmp[:], uint64(k))
_, err = en.Write(tmp[:])
if err != nil {
err = msgp.WrapError(err)
return
}
}
return nil
}