upload_fileobj does not look up previously uploaded parts

[dariusjs]

This is probably an edge case. Initially I tried using the upload_fileobj method within an AWS Lambda function. Some of the downloads I need to perform take longer than the maximum lambda timeout so it ends prematurely. On a next lambda run it ends up creating a new UploadID as evidenced by running the list_parts method.

I have ended up building my own download manager using the available API's that can resume downloads but perhaps it would be worthwhile having the upload_fileobj support this.

Something like this potentially should not sit in a lambda but I would like to see if any others run into this.

[joguSD]

I'm not sure if we would be able to support something like this without making grossly unsafe assumptions about the contents being uploaded. Specifically, we have no way of ensuring that the provided file is going to be the same between two upload_fileobj calls. Does the code you wrote account for this or does it just assume that the contents are the same?

[dariusjs]

No I don’t. I am using dynamodb to check against an md5sum of the file. As I mentioned something like this most likely does not belong in a lambda.