minio/cmd/erasure-common.go
Klaus Post f939d1c183
Independent Multipart Uploads (#15346)
Do completely independent multipart uploads.

In distributed mode, a lock was held to merge each multipart 
upload as it was added. This lock was highly contested and 
retries are expensive (timewise) in distributed mode.

Instead, each part adds its metadata information uniquely. 
This eliminates the per object lock required for each to merge.
The metadata is read back and merged by "CompleteMultipartUpload" 
without locks when constructing final object.

Co-authored-by: Harshavardhana <harsha@minio.io>
2022-07-19 08:35:29 -07:00

207 lines
5.3 KiB
Go

// Copyright (c) 2015-2021 MinIO, Inc.
//
// This file is part of MinIO Object Storage stack
//
// This program is free software: you can redistribute it and/or modify
// it under the terms of the GNU Affero General Public License as published by
// the Free Software Foundation, either version 3 of the License, or
// (at your option) any later version.
//
// This program is distributed in the hope that it will be useful
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
// GNU Affero General Public License for more details.
//
// You should have received a copy of the GNU Affero General Public License
// along with this program. If not, see <http://www.gnu.org/licenses/>.
package cmd
import (
"context"
"fmt"
"sync"
"github.com/minio/minio/internal/logger"
"github.com/minio/minio/internal/sync/errgroup"
)
func (er erasureObjects) getOnlineDisks() (newDisks []StorageAPI) {
disks := er.getDisks()
var wg sync.WaitGroup
var mu sync.Mutex
for _, i := range hashOrder(UTCNow().String(), len(disks)) {
i := i
wg.Add(1)
go func() {
defer wg.Done()
if disks[i-1] == nil {
return
}
di, err := disks[i-1].DiskInfo(context.Background())
if err != nil || di.Healing {
// - Do not consume disks which are not reachable
// unformatted or simply not accessible for some reason.
//
// - Do not consume disks which are being healed
//
// - Future: skip busy disks
return
}
mu.Lock()
newDisks = append(newDisks, disks[i-1])
mu.Unlock()
}()
}
wg.Wait()
return newDisks
}
func (er erasureObjects) getLoadBalancedLocalDisks() (newDisks []StorageAPI) {
disks := er.getDisks()
// Based on the random shuffling return back randomized disks.
for _, i := range hashOrder(UTCNow().String(), len(disks)) {
if disks[i-1] != nil && disks[i-1].IsLocal() {
newDisks = append(newDisks, disks[i-1])
}
}
return newDisks
}
// getLoadBalancedDisks - fetches load balanced (sufficiently randomized) disk slice.
// ensures to skip disks if they are not healing and online.
func (er erasureObjects) getLoadBalancedDisks(optimized bool) []StorageAPI {
disks := er.getDisks()
if !optimized {
var newDisks []StorageAPI
for _, i := range hashOrder(UTCNow().String(), len(disks)) {
newDisks = append(newDisks, disks[i-1])
}
return newDisks
}
var wg sync.WaitGroup
var mu sync.Mutex
newDisks := map[uint64][]StorageAPI{}
// Based on the random shuffling return back randomized disks.
for _, i := range hashOrder(UTCNow().String(), len(disks)) {
i := i
wg.Add(1)
go func() {
defer wg.Done()
if disks[i-1] == nil {
return
}
di, err := disks[i-1].DiskInfo(context.Background())
if err != nil || di.Healing {
// - Do not consume disks which are not reachable
// unformatted or simply not accessible for some reason.
//
// - Do not consume disks which are being healed
//
// - Future: skip busy disks
return
}
mu.Lock()
// Capture disks usage wise upto resolution of MiB
newDisks[di.Used/1024/1024] = append(newDisks[di.Used/1024/1024], disks[i-1])
mu.Unlock()
}()
}
wg.Wait()
var max uint64
for k := range newDisks {
if k > max {
max = k
}
}
// Return disks which have maximum disk usage common.
return newDisks[max]
}
// readMultipleFiles Reads raw data from all specified files from all disks.
func readMultipleFiles(ctx context.Context, disks []StorageAPI, req ReadMultipleReq, readQuorum int) ([]ReadMultipleResp, error) {
resps := make([]chan ReadMultipleResp, len(disks))
for i := range resps {
resps[i] = make(chan ReadMultipleResp, len(req.Files))
}
g := errgroup.WithNErrs(len(disks))
// Read files in parallel across disks.
for index := range disks {
index := index
g.Go(func() (err error) {
if disks[index] == nil {
return errDiskNotFound
}
return disks[index].ReadMultiple(ctx, req, resps[index])
}, index)
}
dataArray := make([]ReadMultipleResp, 0, len(req.Files))
// Merge results. They should come in order from each.
for _, wantFile := range req.Files {
quorum := 0
toAdd := ReadMultipleResp{
Bucket: req.Bucket,
Prefix: req.Prefix,
File: wantFile,
}
for i := range resps {
if disks[i] == nil {
continue
}
select {
case <-ctx.Done():
case gotFile, ok := <-resps[i]:
if !ok {
continue
}
if gotFile.Error != "" || !gotFile.Exists {
continue
}
if gotFile.File != wantFile || gotFile.Bucket != req.Bucket || gotFile.Prefix != req.Prefix {
continue
}
quorum++
if toAdd.Modtime.After(gotFile.Modtime) || len(gotFile.Data) < len(toAdd.Data) {
// Pick latest, or largest to avoid possible truncated entries.
continue
}
toAdd = gotFile
}
}
if quorum < readQuorum {
toAdd.Exists = false
toAdd.Error = errErasureReadQuorum.Error()
toAdd.Data = nil
}
dataArray = append(dataArray, toAdd)
}
errs := g.Wait()
for index, err := range errs {
if err == nil {
continue
}
if !IsErr(err, []error{
errFileNotFound,
errVolumeNotFound,
errFileVersionNotFound,
errDiskNotFound,
errUnformattedDisk,
}...) {
logger.LogOnceIf(ctx, fmt.Errorf("Drive %s, path (%s/%s) returned an error (%w)",
disks[index], req.Bucket, req.Prefix, err),
disks[index].String())
}
}
// Return all the metadata.
return dataArray, nil
}