Verify KestrelServerLimits are enforced for HTTP/3

[halter73]

We should have equivalents to most of Http2TimeoutTests for HTTP/3.

Timeouts to add and/or verify:

• Min combined request body rate timeouts
  • KestrelServerLimits.MinRequestBodyDataRate
  • HTTP/3 should partially support the IHttpMinRequestBodyDataRateFeature.
  • Tracked individually by HTTP/3: Support request body rate timeout #29655
• Min combined response rate timeouts
  • KestrelServerLimits.MinResponseDataRate
  • This is everything sent to the client. The response headers, response body, control frames, etc...
• Response drain timeouts
  • Derived from KestrelServerLimits.MinResponseDataRate and KestrelServerLimits.MaxResponseBufferSize
  • When there's backpressure on the underlying stream/connection after PipeWriter.FlushAsync() completes which can happen with non-zero PauseWriterThreshold.
• Request header timeouts
  • KestrelServerLimits.RequestHeadersTimeout
• Keep-alive timeouts (will use msquic timeout)
  • KestrelServerLimits.KeepAliveTimeout
  • We probably don't want to keep idle QuicConnections around forever

If any the above bullet points aren't implemented yet, they probably deserve their own issue. The same goes for the size and count based limits in KestrelServerLimits:

• MaxRequestLineSize
• MaxRequestHeadersTotalSize
• MaxRequestHeaderCount
• MaxRequestBodySize (+IHttpMaxRequestBodySizeFeature)
• MaxConcurrentConnections

[ghost]

Thanks for contacting us.
We're moving this issue to the Next sprint planning milestone for future evaluation / consideration. We will evaluate the request when we are planning the work for the next milestone. To learn more about what to expect next and how this issue will be handled you can read more about our triage process here.

[BrennanConroy]

@wtgodbe Would you be interested in starting on this for preview3?

[JamesNK]

This is pretty big. I'm going to start on the first two: min request and response rates.

[wtgodbe]

I'd be happy to help out here - @JamesNK I'll reach out to you once I have time to see where I could be most useful

[JamesNK]

re: RequestHeadersTimeout

Decision is for now HTTP/3 won't use it. In HTTP/2 RequestHeadersTimeout is used as a timeout between receiving the HEADER frame and any CONTINUATION frames at the start of the request. In HTTP/3 the HEADERS frame always contains all HEADERS, so the timeout no longer has the same use in HTTP/3.

Will use the timeout. Timeout starts before the request stream starts awaiting data and is canceled once HEADERS frame is received.