[TODO] Explore handling HTTP errors on Rektor flakiness

[webknjaz]

@woodruffw @facutuesca we recently saw an HTTP 502 and a traceback in the attestations flow:

Traceback (most recent call last):
  File "/root/.local/lib/python3.12/site-packages/sigstore/_internal/rekor/client.py", line 160, in post
    resp.raise_for_status()
  File "/root/.local/lib/python3.12/site-packages/requests/models.py", line 1024, in raise_for_status
    raise HTTPError(http_error_msg, response=self)
requests.exceptions.HTTPError: 502 Server Error: Bad Gateway for url: https://rekor.sigstore.dev/api/v1/log/entries/

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/app/attestations.py", line 149, in <module>
    main()
  File "/app/attestations.py", line 145, in main
    attest_dist(dist_path, attestation_path, signer)
  File "/app/attestations.py", line 114, in attest_dist
    attestation = Attestation.sign(signer, dist)
                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/root/.local/lib/python3.12/site-packages/pypi_attestations/_impl.py", line 200, in sign
    bundle = signer.sign_dsse(stmt)
             ^^^^^^^^^^^^^^^^^^^^^^
  File "/root/.local/lib/python3.12/site-packages/sigstore/sign.py", line 230, in sign_dsse
    return self._finalize_sign(cert, content, proposed_entry)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/root/.local/lib/python3.12/site-packages/sigstore/sign.py", line 189, in _finalize_sign
    entry = self._signing_ctx._rekor.log.entries.post(proposed_entry)
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/root/.local/lib/python3.12/site-packages/sigstore/_internal/rekor/client.py", line 162, in post
    raise RekorClientError(http_error)
sigstore._internal.rekor.client.RekorClientError: Rekor returned an unknown error with HTTP 502

(https://github.com/aio-libs/aiohttp/actions/runs/15359675323/job/43225662768#step:9:384)

Mind taking a look?

[webknjaz]

Maybe, this should be fixed in the sigstore lib. Not sure. If not, I suspect Twine would need some handling as well.

[woodruffw]

Yeah, it looks like there was some kind of Rekor hiccup/short outage over the weekend -- pinging @haydentherapper since he might know more 🙂

Looks like there are two things sigstore-python could do better here:

• We could probably produce a more explanatory error here (both for API and CLI users)
• We should probably have some kind of retry handling for Rekor API calls, although in this case that probably wouldn't have helped much

[webknjaz]

yeah, that's exactly what I was thinking

[haydentherapper]

Yea, there was a brief outage over the weekend, still investigating root cause.