speed up startup sequence for all operations (#14148)

This speed-up is intended for faster startup times
for almost all MinIO operations. Changes here are

- Drives are not re-read for 'format.json' on a regular
  basis once read during init is remembered and refreshed
  at 5 second intervals.

- Do not do O_DIRECT tests on drives with existing 'format.json'
  only fresh setups need this check.

- Parallelize initializing erasureSets for multiple sets.

- Avoid re-reading format.json when migrating 'format.json'
  from really old V1->V2->V3

- Keep a copy of local drives for any given server in memory
  for a quick lookup.
This commit is contained in:
Harshavardhana
2022-01-24 11:28:45 -08:00
committed by GitHub
parent f30afa4956
commit 5a9f133491
14 changed files with 308 additions and 281 deletions

View File

@@ -110,6 +110,8 @@ type xlStorage struct {
diskInfoCache timedValue
sync.RWMutex
formatData []byte
// mutex to prevent concurrent read operations overloading walks.
walkMu sync.Mutex
walkReadMu sync.Mutex
@@ -205,23 +207,9 @@ func newLocalXLStorage(path string) (*xlStorage, error) {
})
}
// Sanitize - sanitizes the `format.json`, cleanup tmp.
// all other future cleanups should be added here.
func (s *xlStorage) Sanitize() error {
if err := formatErasureMigrate(s.diskPath); err != nil && !errors.Is(err, os.ErrNotExist) {
return err
}
// Create any missing paths.
makeFormatErasureMetaVolumes(s)
return formatErasureCleanupTmp(s.diskPath)
}
// Initialize a new storage disk.
func newXLStorage(ep Endpoint) (*xlStorage, error) {
func newXLStorage(ep Endpoint) (s *xlStorage, err error) {
path := ep.Path
var err error
if path, err = getValidPath(path); err != nil {
return nil, err
}
@@ -255,7 +243,7 @@ func newXLStorage(ep Endpoint) (*xlStorage, error) {
}
}
p := &xlStorage{
s = &xlStorage{
diskPath: path,
endpoint: ep,
globalSync: env.Get(config.EnvFSOSync, config.EnableOff) == config.EnableOn,
@@ -265,42 +253,56 @@ func newXLStorage(ep Endpoint) (*xlStorage, error) {
diskIndex: -1,
}
go formatErasureCleanupTmp(s.diskPath) // cleanup any old data.
formatData, formatFi, err := formatErasureMigrate(s.diskPath)
if err != nil && !errors.Is(err, os.ErrNotExist) {
if os.IsPermission(err) {
return nil, errDiskAccessDenied
} else if isSysErrIO(err) {
return nil, errFaultyDisk
}
return nil, err
}
s.formatData = formatData
s.formatFileInfo = formatFi
if len(s.formatData) == 0 { // Unformatted disk check if O_DIRECT is supported.
// Check if backend is writable and supports O_DIRECT
var rnd [32]byte
_, _ = rand.Read(rnd[:])
filePath := pathJoin(s.diskPath, ".writable-check-"+hex.EncodeToString(rnd[:])+".tmp")
w, err := s.openFileDirect(filePath, os.O_CREATE|os.O_WRONLY|os.O_EXCL)
if err != nil {
return s, err
}
_, err = w.Write(alignedBuf)
w.Close()
if err != nil {
if isSysErrInvalidArg(err) {
return s, errUnsupportedDisk
}
return s, err
}
Remove(filePath)
} else {
format := &formatErasureV3{}
json := jsoniter.ConfigCompatibleWithStandardLibrary
if err = json.Unmarshal(s.formatData, &format); err != nil {
return s, errCorruptedFormat
}
s.diskID = format.Erasure.This
s.formatLastCheck = time.Now()
s.formatLegacy = format.Erasure.DistributionAlgo == formatErasureVersionV2DistributionAlgoV1
}
// Create all necessary bucket folders if possible.
if err = p.MakeVolBulk(context.TODO(), minioMetaBucket, minioMetaTmpBucket, minioMetaMultipartBucket, dataUsageBucket, minioMetaSpeedTestBucket); err != nil {
if err = makeFormatErasureMetaVolumes(s); err != nil {
return nil, err
}
// Check if backend is writable and supports O_DIRECT
var rnd [8]byte
_, _ = rand.Read(rnd[:])
tmpFile := ".writable-check-" + hex.EncodeToString(rnd[:]) + ".tmp"
filePath := pathJoin(p.diskPath, minioMetaTmpBucket, tmpFile)
w, err := OpenFileDirectIO(filePath, os.O_CREATE|os.O_WRONLY|os.O_EXCL, 0o666)
if err != nil {
switch {
case isSysErrInvalidArg(err):
return p, errUnsupportedDisk
case osIsPermission(err):
return p, errDiskAccessDenied
case isSysErrIO(err):
return p, errFaultyDisk
case isSysErrNotDir(err):
return p, errDiskNotDir
}
return p, err
}
if _, err = w.Write(alignedBuf); err != nil {
w.Close()
if isSysErrInvalidArg(err) {
return p, errUnsupportedDisk
}
return p, err
}
w.Close()
Remove(filePath)
// Success.
return p, nil
return s, nil
}
// getDiskInfo returns given disk information.
@@ -308,7 +310,6 @@ func getDiskInfo(diskPath string) (di disk.Info, err error) {
if err = checkPathLength(diskPath); err == nil {
di, err = disk.GetInfo(diskPath)
}
switch {
case osIsNotExist(err):
err = errDiskNotFound
@@ -371,7 +372,7 @@ func (s *xlStorage) SetDiskLoc(poolIdx, setIdx, diskIdx int) {
func (s *xlStorage) Healing() *healingTracker {
healingFile := pathJoin(s.diskPath, minioMetaBucket,
bucketMetaPrefix, healingTrackerFilename)
b, err := xioutil.ReadFile(healingFile)
b, err := ioutil.ReadFile(healingFile)
if err != nil {
return nil
}
@@ -624,8 +625,8 @@ func (s *xlStorage) GetDiskID() (string, error) {
fileInfo := s.formatFileInfo
lastCheck := s.formatLastCheck
// check if we have a valid disk ID that is less than 1 second old.
if fileInfo != nil && diskID != "" && time.Since(lastCheck) <= time.Second {
// check if we have a valid disk ID that is less than 1 seconds old.
if fileInfo != nil && diskID != "" && time.Since(lastCheck) <= 1*time.Second {
s.RUnlock()
return diskID, nil
}
@@ -645,7 +646,7 @@ func (s *xlStorage) GetDiskID() (string, error) {
}
formatFile := pathJoin(s.diskPath, minioMetaBucket, formatConfigFile)
b, err := xioutil.ReadFile(formatFile)
b, err := ioutil.ReadFile(formatFile)
if err != nil {
// If the disk is still not initialized.
if osIsNotExist(err) {
@@ -676,6 +677,7 @@ func (s *xlStorage) GetDiskID() (string, error) {
s.Lock()
defer s.Unlock()
s.formatData = b
s.diskID = format.Erasure.This
s.formatLegacy = format.Erasure.DistributionAlgo == formatErasureVersionV2DistributionAlgoV1
s.formatFileInfo = fi
@@ -1417,6 +1419,16 @@ func (s *xlStorage) readAllData(ctx context.Context, volumeDir string, filePath
// This API is meant to be used on files which have small memory footprint, do
// not use this on large files as it would cause server to crash.
func (s *xlStorage) ReadAll(ctx context.Context, volume string, path string) (buf []byte, err error) {
// Specific optimization to avoid re-read from the drives for `format.json`
// in-case the caller is a network operation.
if volume == minioMetaBucket && path == formatConfigFile {
s.RLock()
formatData := s.formatData
s.RUnlock()
if len(formatData) > 0 {
return formatData, nil
}
}
volumeDir, err := s.getVolDir(volume)
if err != nil {
return nil, err
@@ -1537,6 +1549,30 @@ func (s *xlStorage) ReadFile(ctx context.Context, volume string, path string, of
return int64(len(buffer)), nil
}
func (s *xlStorage) openFileDirect(path string, mode int) (f *os.File, err error) {
// Create top level directories if they don't exist.
// with mode 0o777 mkdir honors system umask.
mkdirAll(pathutil.Dir(path), 0o777) // don't need to fail here
w, err := OpenFileDirectIO(path, mode, 0o666)
if err != nil {
switch {
case isSysErrInvalidArg(err):
return nil, errUnsupportedDisk
case osIsPermission(err):
return nil, errDiskAccessDenied
case isSysErrIO(err):
return nil, errFaultyDisk
case isSysErrNotDir(err):
return nil, errDiskNotDir
case os.IsNotExist(err):
return nil, errDiskNotFound
}
}
return w, nil
}
func (s *xlStorage) openFileSync(filePath string, mode int) (f *os.File, err error) {
// Create top level directories if they don't exist.
// with mode 0777 mkdir honors system umask.