#415 feat: Add retry logic to file upload function Pull request opened by pingsutw This pull request introduces a robust retry mechanism for file uploads in the

src/flyte/remote/_data.py

module. The main improvement is the addition of an exponential backoff strategy for handling transient network and server errors during file uploads, replacing the previous single-attempt upload logic. This enhances reliability, especially in unstable network conditions. Resilience improvements for file uploads: • Added a new

_upload_with_retry

async function that uploads files using exponential backoff and retries on transient network errors and retryable HTTP status codes (408, 429, 500, 502, 503, 504). This function logs retries and raises a

RuntimeSystemError

if all attempts fail. • Introduced a custom

_RetryableUploadError

exception and a helper function

_is_retryable_status_code

to distinguish between retryable and non-retryable HTTP errors. Integration with file upload flow: • Updated the

_upload_single_file

function to use

_upload_with_retry

instead of directly uploading the file in a single attempt, ensuring all uploads benefit from the new retry logic. • Refactored header preparation and removed redundant file open/close logic in

_upload_single_file

, delegating the upload responsibility to the new retry-enabled function. flyteorg/flyte-sdk