Jacob Bohanon

``` Step #12 - "test": [2022-09-16 18:57:29.569][86309][critical][assert] [external/envoy/source/extensions/filters/common/ratelimit/ratelimit_impl.cc:77] assert failure: response->overall_code() != envoy::service::ratelimit::v3::RateLimitResponse::UNKNOWN. ``` This line is the culprit. @Bslabe123 and I thought this was fixed by changing the rate-limiter...

The fix was released in `0.7.4` of the rate-limiter, and 1.12 (where this flake occurred) is on `0.7.3`

It appears this is an upstream bug that has existed for quite some time. All versions I tested from 1.24 back to 1.17 exhibited the errant behavior. The following is...

@yuval-k Is there anything in here that rings any bells? @nfuden mentioned you had experience with similar bugs in upstream.

@kevin-shelaga I have created upstream bug-report issue [envoyproxy/envoy#23625](https://github.com/envoyproxy/envoy/issues/23625). Please let us know how to prioritize the further investigation into the upstream bug.

The following diff should be applied when Gloo Edge 1.17.0 is released in order to fix the string matching ```diff diff --git a/Formula/g/glooctl.rb b/Formula/g/glooctl.rb index 37d91cc4eab..f6e98695a42 100644 --- a/Formula/g/glooctl.rb +++...

I have confirmed that the homebrew bumps are handled by the homebrew repo and not by our code anymore. The GitHub Action is [here](https://github.com/Homebrew/homebrew-core/blob/8064f66cd04d0f32dc1be25ce8363a7a9e370fae/.github/workflows/autobump.yml) and `glooctl` is on the autobump...

~~Can you tell from the testing if this behavior driven by the ordering of the `setDescriptors` definition, or if it is by the descriptor nearest being limited?~~ ~~It's not really...

almost everything about my last comment was incorrect: - We return all of the descriptor statuses to Envoy - Envoy determines the descriptor that is nearest to being limited and...

[RFC spec](https://www.rfc-editor.org/rfc/rfc9110.html#section-5.3-2) indicates comma-separated multi-value headers *MUST NOT* have their order changed, so we could feasibly propose an option to return all descriptors in the `X-RateLimit` headers if absolutely necessary,...