SG Ground Truth

multipart lifetime

An upload_id still completed 1230 s after its init, with part 2 minted then. The link's timestamp is not age-checked: it is part of the stored key, so another value signs a URL S3 answers 404.

Partly measured. an upload_id older than 1230 s. Blocked on time: `139 --write --wait <seconds>` measures any age

API

Q How long does a multipart upload_id live, is the timestamp in a part link checked for age, and does a part URL minted 20 minutes after the init still complete?

Endpoint GET /entity/<type>/<id>/_upload?multipart_upload=true ; GET .../_upload/multipart ; PUT <links.upload> ; POST <links.complete_upload>

Docs claim Silent on lifetimes. Probe 044 measured the flow in one sitting, probe 125 a presigned URL refused after 900 s.

Actual

M1 init t0, timestamp 2026-11-11T19:07:29Z; get_next_part query: filename part_number timestamp upload_id upload_type
  part 1 (5 MiB) PUT at t0+3s                  -> 200, ETag
  part 1 URL PUT again at t0+1230s             -> 403 AccessDenied "Request has expired", X-Amz-Expires 900
  get_next_part at t0+1231s                    -> 200, a fresh part 2 URL
  part 2 PUT                                   -> 200
  complete with both ETags at t0+1232s         -> 201; the stored file reads back

M2 part 1 URL minted by GET .../_upload/multipart, the init's query with timestamp replaced
  2020-01-01T00:00:00Z  -> 200; URL path differs from the init's; X-Amz-Date the call's own time
                           PUT -> 404 <Code>NoSuchUpload</Code> "The specified upload does not exist.
                                  The upload ID may be invalid, or the upload may have been aborted or completed."
  zz                    -> 400 {"status": "error", "error": "no time information in \"zz UTC\""}
  omitted               -> 400 code 103 "Request Parameters invalid." source {"timestamp": ["timestamp is missing"]}
  the init's own value  -> 200; same path as the init's URL; PUT 200; complete 201

Teaches

  • A resumed upload keeps its upload_id, timestamp and the ETags of finished parts, and asks get_next_part for a fresh URL per part: every URL dies 900 s after it was minted (probe 125), the upload_id did not within 1230 s. Persist the init's whole data object with the ETags.
  • timestamp is an identifier, not a freshness check. A wrong value is accepted at minting and fails only at the part PUT, as an S3 404 that blames the upload ID. Send the init's value verbatim.
  • A garbage timestamp answers a bare {"status": "error", ...} body outside the errors[] envelope; parse both shapes on this call.
  • An init that is never completed nor aborted stays open on storage (probe 044). Its lifetime there is the storage's and was not measured past 1230 s.

Every entry on this site is the output of a probe in probes/. The corpus is generated by running those probes against a live Flow Production Tracking site, not written from memory.

Not affiliated with or endorsed by Autodesk. Flow Production Tracking is their product; this is an independent record of how its REST API answers.