Files
rclone/backend/box/upload.go
Nick Craig-Wood fb8783732d box: reuse multipart upload part buffers via the global pool
Each part of a multipart upload allocated a fresh part-sized buffer
(the size is chosen by Box, typically 8-32 MiB) with up to --transfers
parts in flight, so large uploads churned allocations and GC.

Buffer parts with multipart.NewRW instead so the memory comes from
rclone's global pool, is reused across parts and files, and is part of
rclone's central memory management.

The pool.RW is seekable so the retry closure seeks back to the start
before each attempt instead of rebuilding a bytes.Reader, and the
per-part SHA1 digest is computed by reading the buffer and seeking
back. The body goes through lib/rest which already stops the transport
from closing it; the uploading goroutine owns and closes the buffer.
The whole-file SHA1 used for the commit is unchanged.

The FsPutRetry integration test covers the retry of a failed upload
request and checks the buffers are returned to the pool.
2026-09-01 14:21:52 +01:00

282 lines
8.1 KiB
Go

// multipart upload for box
package box
import (
"context"
"crypto/sha1"
"encoding/base64"
"encoding/json"
"errors"
"fmt"
"io"
"net/http"
"strconv"
"sync"
"time"
"github.com/rclone/rclone/backend/box/api"
"github.com/rclone/rclone/fs"
"github.com/rclone/rclone/fs/accounting"
"github.com/rclone/rclone/lib/atexit"
"github.com/rclone/rclone/lib/multipart"
"github.com/rclone/rclone/lib/rest"
)
// createUploadSession creates an upload session for the object
func (o *Object) createUploadSession(ctx context.Context, leaf, directoryID string, size int64) (response *api.UploadSessionResponse, err error) {
opts := rest.Opts{
Method: "POST",
Path: "/files/upload_sessions",
RootURL: uploadURL,
}
request := api.UploadSessionRequest{
FileSize: size,
}
// If object has an ID then it is existing so create a new version
if o.id != "" {
opts.Path = "/files/" + o.id + "/upload_sessions"
} else {
opts.Path = "/files/upload_sessions"
request.FolderID = directoryID
request.FileName = o.fs.opt.Enc.FromStandardName(leaf)
}
var resp *http.Response
err = o.fs.pacer.Call(func() (bool, error) {
resp, err = o.fs.srv.CallJSON(ctx, &opts, &request, &response)
return shouldRetry(ctx, resp, err)
})
return
}
// sha1Digest produces a digest using sha1 as per RFC3230
func sha1Digest(digest []byte) string {
return "sha=" + base64.StdEncoding.EncodeToString(digest)
}
// uploadPart uploads a part in an upload session
//
// chunk must contain exactly the part's data and is read from the start
// on each attempt.
func (o *Object) uploadPart(ctx context.Context, SessionID string, offset, totalSize int64, chunk io.ReadSeeker, wrap accounting.WrapFn, options ...fs.OpenOption) (response *api.UploadPartResponse, err error) {
hasher := sha1.New()
chunkSize, err := io.Copy(hasher, chunk)
if err != nil {
return nil, err
}
opts := rest.Opts{
Method: "PUT",
Path: "/files/upload_sessions/" + SessionID,
RootURL: uploadURL,
ContentType: "application/octet-stream",
ContentLength: &chunkSize,
ContentRange: fmt.Sprintf("bytes %d-%d/%d", offset, offset+chunkSize-1, totalSize),
Options: options,
ExtraHeaders: map[string]string{
"Digest": sha1Digest(hasher.Sum(nil)),
},
}
var resp *http.Response
err = o.fs.pacer.Call(func() (bool, error) {
_, err = chunk.Seek(0, io.SeekStart)
if err != nil {
return false, err
}
opts.Body = wrap(chunk)
resp, err = o.fs.srv.CallJSON(ctx, &opts, nil, &response)
return shouldRetry(ctx, resp, err)
})
if err != nil {
return nil, err
}
return response, nil
}
// commitUpload finishes an upload session
func (o *Object) commitUpload(ctx context.Context, SessionID string, parts []api.Part, modTime time.Time, sha1sum []byte) (result *api.FolderItems, err error) {
opts := rest.Opts{
Method: "POST",
Path: "/files/upload_sessions/" + SessionID + "/commit",
RootURL: uploadURL,
ExtraHeaders: map[string]string{
"Digest": sha1Digest(sha1sum),
},
}
request := api.CommitUpload{
Parts: parts,
}
request.Attributes.ContentModifiedAt = api.Time(modTime)
request.Attributes.ContentCreatedAt = api.Time(modTime)
var body []byte
var resp *http.Response
// For discussion of this value see:
// https://github.com/rclone/rclone/issues/2054
maxTries := o.fs.opt.CommitRetries
const defaultDelay = 10
var tries int
outer:
for tries = range maxTries {
err = o.fs.pacer.Call(func() (bool, error) {
resp, err = o.fs.srv.CallJSON(ctx, &opts, &request, nil)
if err != nil {
return shouldRetry(ctx, resp, err)
}
body, err = rest.ReadBody(resp)
return shouldRetry(ctx, resp, err)
})
delay := defaultDelay
var why string
if err != nil {
// Sometimes we get 400 Error with
// parts_mismatch immediately after uploading
// the last part. Ignore this error and wait.
if boxErr, ok := err.(*api.Error); ok && boxErr.Code == "parts_mismatch" {
why = err.Error()
} else {
return nil, err
}
} else {
switch resp.StatusCode {
case http.StatusOK, http.StatusCreated:
break outer
case http.StatusAccepted:
why = "not ready yet"
delayString := resp.Header.Get("Retry-After")
if delayString != "" {
delay, err = strconv.Atoi(delayString)
if err != nil {
fs.Debugf(o, "Couldn't decode Retry-After header %q: %v", delayString, err)
delay = defaultDelay
}
}
default:
return nil, fmt.Errorf("unknown HTTP status return %q (%d)", resp.Status, resp.StatusCode)
}
}
fs.Debugf(o, "commit multipart upload failed %d/%d - trying again in %d seconds (%s)", tries+1, maxTries, delay, why)
time.Sleep(time.Duration(delay) * time.Second)
}
if tries >= maxTries {
return nil, errors.New("too many tries to commit multipart upload - increase --low-level-retries")
}
err = json.Unmarshal(body, &result)
if err != nil {
return nil, fmt.Errorf("couldn't decode commit response: %q: %w", body, err)
}
return result, nil
}
// abortUpload cancels an upload session
func (o *Object) abortUpload(ctx context.Context, SessionID string) (err error) {
opts := rest.Opts{
Method: "DELETE",
Path: "/files/upload_sessions/" + SessionID,
RootURL: uploadURL,
NoResponse: true,
}
var resp *http.Response
err = o.fs.pacer.Call(func() (bool, error) {
resp, err = o.fs.srv.Call(ctx, &opts)
return shouldRetry(ctx, resp, err)
})
return err
}
// uploadMultipart uploads a file using multipart upload
func (o *Object) uploadMultipart(ctx context.Context, in io.Reader, leaf, directoryID string, size int64, modTime time.Time, options ...fs.OpenOption) (err error) {
// Create upload session
session, err := o.createUploadSession(ctx, leaf, directoryID, size)
if err != nil {
return fmt.Errorf("multipart upload create session failed: %w", err)
}
chunkSize := session.PartSize
fs.Debugf(o, "Multipart upload session started for %d parts of size %v", session.TotalParts, fs.SizeSuffix(chunkSize))
// Cancel the session if something went wrong
defer atexit.OnError(&err, func() {
fs.Debugf(o, "Cancelling multipart upload: %v", err)
cancelErr := o.abortUpload(ctx, session.ID)
if cancelErr != nil {
fs.Logf(o, "Failed to cancel multipart upload: %v", cancelErr)
}
})()
// unwrap the accounting from the input, we use wrap to put it
// back on after the buffering
in, wrap := accounting.UnWrap(in)
// Upload the chunks
remaining := size
position := int64(0)
parts := make([]api.Part, session.TotalParts)
hash := sha1.New()
errs := make(chan error, 1)
var wg sync.WaitGroup
outer:
for part := range session.TotalParts {
// Check any errors
select {
case err = <-errs:
break outer
default:
}
reqSize := min(remaining, chunkSize)
// Read the chunk into memory from the global pool, hashing
// it for the whole file hash (must be done sequentially)
rw := multipart.NewRW().Reserve(reqSize)
_, err = io.CopyN(rw, io.TeeReader(in, hash), reqSize)
if err != nil {
_ = rw.Close()
err = fmt.Errorf("multipart upload failed to read source: %w", err)
break outer
}
// Transfer the chunk
wg.Add(1)
o.fs.uploadToken.Get()
go func(part int, position int64) {
defer wg.Done()
defer o.fs.uploadToken.Put()
defer func() { _ = rw.Close() }()
fs.Debugf(o, "Uploading part %d/%d offset %v/%v part size %v", part+1, session.TotalParts, fs.SizeSuffix(position), fs.SizeSuffix(size), fs.SizeSuffix(chunkSize))
partResponse, err := o.uploadPart(ctx, session.ID, position, size, rw, wrap, options...)
if err != nil {
err = fmt.Errorf("multipart upload failed to upload part: %w", err)
select {
case errs <- err:
default:
}
return
}
parts[part] = partResponse.Part
}(part, position)
// ready for next block
remaining -= chunkSize
position += chunkSize
}
wg.Wait()
if err == nil {
select {
case err = <-errs:
default:
}
}
if err != nil {
return err
}
// Finalise the upload session
result, err := o.commitUpload(ctx, session.ID, parts, modTime, hash.Sum(nil))
if err != nil {
return fmt.Errorf("multipart upload failed to finalize: %w", err)
}
if result.TotalCount != 1 || len(result.Entries) != 1 {
return fmt.Errorf("multipart upload failed %v - not sure why", o)
}
return o.setMetaData(&result.Entries[0])
}