Add ExecutionTimeout parameter

[benaadams]

Timeout between two successive body reads which if exceeded the client is determined to be disconnected and the request aborted.

Failure state: 408 Request Timeout

ngnix default client_body_timeout 60s

Implementation-wise the timer can use the same timer as Keep-alive; as a multi stage/state timer.

When deployed in recommended configuration with a reverse proxy (e.g. IIS/nginx) this would be handled by the proxy - however, this wouldn't apply in a pure Kestrel deployment.

[Tratcher]

Make sure this doesn't break websockets.

[benaadams]

Re: websockets @Tratcher I don't think so; but https isn't an upgrade request is it?

[Tratcher]

https isn't an upgrade request is it?

Huh?

[benaadams]

Huh?

My plan is to switch off the timers on opaque stream upgrade and assume they were handling that interaction (would also need to be same for MaxUploadBytes) - just making sure there is no corner cases I hadn't thought of (https in this case)

[Tratcher]

Ah, sounds good.

https shouldn't be any different from http in this scenario.

[benaadams]

Would be easier to have this as a maxRequestExecutionTimeout, that includes body read time, that work?

Would probably also be more useful that way.

/cc @blowdart @halter73 @CesarBS @muratg

[benaadams]

Hmmm... may cause issue with large file uploads; so may still need a sliding window for body requests :-/

[blowdart]

Hurrah, timeouts :D

[DamianEdwards]

post RC-2