upload failed after %d retries due to server errors, error %

Error message

upload failed after %d retries due to server errors, error %d

What it means

Fatal retry exhaustion in Cloudreve's OneDrive-mediated upload (upOneDrive). Chunks are PUT directly to the OneDrive upload session URL; HTTP 500-504 responses are retried with exponential backoff. When retries exceed maxRetries, the loop aborts with this message, embedding the last HTTP status code.

Source

Thrown at drivers/cloudreve/util.go:348

			return err
		}
		req = req.WithContext(ctx)
		req.ContentLength = byteSize
		// req.Header.Set("Content-Length", strconv.Itoa(int(byteSize)))
		req.Header.Set("Content-Range", fmt.Sprintf("bytes %d-%d/%d", finish, finish+byteSize-1, stream.GetSize()))
		req.Header.Set("User-Agent", d.getUA())
		finish += byteSize
		res, err := base.HttpClient.Do(req)
		if err != nil {
			return err
		}
		// https://learn.microsoft.com/zh-cn/onedrive/developer/rest-api/api/driveitem_createuploadsession
		switch {
		case res.StatusCode >= 500 && res.StatusCode <= 504:
			retryCount++
			if retryCount > maxRetries {
				res.Body.Close()
				return fmt.Errorf("upload failed after %d retries due to server errors, error %d", maxRetries, res.StatusCode)
			}
			backoff := time.Duration(1<<retryCount) * time.Second
			utils.Log.Warnf("[Cloudreve-OneDrive] server errors %d while uploading, retrying after %v...", res.StatusCode, backoff)
			time.Sleep(backoff)
		case res.StatusCode != 201 && res.StatusCode != 202 && res.StatusCode != 200:
			data, _ := io.ReadAll(res.Body)
			res.Body.Close()
			return errors.New(string(data))
		default:
			res.Body.Close()
			retryCount = 0
			finish += byteSize
			up(float64(finish) * 100 / float64(stream.GetSize()))
		}
	}
	// 上传成功发送回调请求
	return d.request(http.MethodPost, "/callback/onedrive/finish/"+u.SessionID, func(req *resty.Request) {
		req.SetBody("{}")

View on GitHub (pinned to 843d9dc814)

Solutions

Retry the whole upload after a wait — transient OneDrive 5xx storms usually clear
Increase proxy read/write timeouts (or bypass the proxy for upload URLs) so chunk PUTs are not converted into 504s
Check OneDrive service status and the Cloudreve session's validity; restart the upload to get a fresh session if it expired
Raise maxRetries or use smaller chunks so each retry window is shorter

Defensive patterns

Strategy: retry

Validate before calling

// pre-flight: confirm the OneDrive session URL is alive with a cheap 0-byte request if supported, and check proxy timeouts > chunk transfer time

Try / catch

if err := d.upOneDrive(ctx, stream, u, up); err != nil {
    if strings.Contains(err.Error(), "retries due to server errors") {
        // session may be stale or OneDrive degraded: get a fresh session and retry once
        u2, err2 := d.getUploadInfo(ctx, stream)
        if err2 == nil {
            err = d.upOneDrive(ctx, stream, u2, up)
        }
    }
}

Prevention

Set reverse-proxy timeouts above worst-case chunk upload duration
Retry whole uploads (fresh session) rather than extending per-chunk retries indefinitely
Watch OneDrive status pages during incident windows and defer bulk uploads

When it happens

Trigger: Every chunk PUT to the OneDrive upload URL returns a 5xx status (500-504) more than maxRetries times consecutively: OneDrive service degradation, the upload session expired server-side (some 5xx masks are actually session-invalid), or an intermediary proxy manufacturing 5xx responses.

Common situations: OneDrive outages or throttling during large uploads; a reverse proxy in front of Cloudreve timing out and returning 502/504 per chunk; slow networks where each chunk exceeds proxy timeouts.

Related errors

AI-assisted analysis of AlistGo/alist@843d9dc814 (2026-08-15). Data as JSON: /api/errors/92ba08455d521f28. Report an issue: GitHub.