Why there is a partNumber in GetObjectRequest?

[crusader-mike]

It seems that you can specify partNumber in GetObject request. What is the associated semantic -- can I download object piece-by-piece using this?

[marcomagdy]

Yes, you can.

[crusader-mike]

So, which byte range GetObjectRequest with partNumber =5 will retrieve? Where can I get such information (along with total number of parts)? I can't find anything in documentation...

[marcomagdy]

This applies for multi-part uploads only. Specifically, multipart uploads that have neither completed nor aborted.
Essentially, this allows you to download a part of an object you're uploading via multi-part upload.

See http://docs.aws.amazon.com/AmazonS3/latest/dev/mpuoverview.html

[crusader-mike]

I see... Thank you.

Is there any way to implement multi-part download (that is download different parts of the same object in parallel) besides using GetObject with byte ranges?

[marcomagdy]

I don't think so. Range-get is the only way I'm aware of.

[vlovich]

To anyone else coming across this link, the explanation doesn't seem correct. AFAICT there's no way to access the contents of an in-progress upload. This should be pretty clear because the GetObject API does not take in an upload ID. So if a part number were referring to an incomplete/not aborted uploaded, which upload would that be? There could be several in flight at one time after all.

The part number in GetObject actually corresponds to the byte range of a published multipart upload. So if you published parts 1(10MB), 2(10MB), and 3(3MB) and ask for part 2, you'll get the byte range [10MB, 20MB).

What's not clear to me is what happens if you upload parts 1 3 & 7 and then ask for part 3. Do you get back part #3 that you uploaded or the third consecutive part contained within the file (i.e. part 7 in the upload). It's not hard to test but I'm lazy.