birenroy

/assign @diannahu /assign @RyanTheOptimist

> Is this a real failure, or a flake? > > ``` > [ FAILED ] IpVersions/Http2FrameIntegrationTest.AdjustUpstreamSettingsMaxStreams/IPv6_Nghttp2, where GetParam() = 8-byte object > ``` > > https://dev.azure.com/cncf/envoy/_build/results?buildId=171029&view=logs&jobId=767be981-567e-57d8-68c3-2140ede0a0bd&j=767be981-567e-57d8-68c3-2140ede0a0bd&t=2181edf2-f610-59f2-c43a-04bb9d0bca00 Test passes with...

/retest

I'd like to confirm my understanding of the problem. Clients speaking HTTP/2 send POST requests to Envoy. Since the upstream_request_attempt_count is nonzero, it seems like the Envoy is forwarding request...

Could you also provide the [initial stream and connection flow control window sizes](https://github.com/envoyproxy/envoy/blob/main/api/envoy/config/core/v3/protocol.proto#L478-L484) that your Envoy is using? The symptoms sound like a possible flow control issue, but the flow...

We identified a potential flow control leak in oghttp2. Please stand by for a fix.

> We saw connection hang in #31710 as well. Maybe the fix was not backported to 1.29? The fix for #31710 was included in v1.29. This looks like a separate...

We identified and fixed a slow flow control leak, which could be the cause of the hangs you experienced. If you're able to pull the latest revision of v1.29, please...

> Can you help give as an idea of what type of request would trigger this issue? The fix in https://github.com/google/quiche/commit/9e875938052443c7f51b25012093be59e20598ac relates to the handling of HTP/2 `DATA` frames with...

I think we can mark this closed until we have evidence of connection hangs unrelated to padding or flow control.