W3C home > Mailing lists > Public > ietf-http-wg@w3.org > April to June 2014

Re: Graceful shutdown #458

From: Roberto Peon <grmocg@gmail.com>
Date: Thu, 8 May 2014 15:01:11 -0700
Message-ID: <CAP+FsNephB9sdsqLV9O8h-O3aaA_PQvfUPrtswXU5KJW6ySiVA@mail.gmail.com>
To: Martin Thomson <martin.thomson@gmail.com>
Cc: HTTP Working Group <ietf-http-wg@w3.org>
DRAINING differs from ALTSVC mostly in that it requires no state be kept.
Otherwise, I believe ALTSVC can be used to drain things so long as it is
worded such that an ALTSVC to the same location causes a new connection to
be created.

-=R


On Thu, May 8, 2014 at 2:55 PM, Martin Thomson <martin.thomson@gmail.com>wrote:

> The shutting down state machine currently looks something like this:
>
>   accepting requests
>     --[GOAWAY]-> not accepting requests
>     --[TCP FIN]-> closed
>
> And during the "not accepting requests" state, it may be that some
> requests are lost.  You will have to check that each request made it.
>
> I proposed (and Daniel provided pull #475) to not change this state
> machine, but to exploit the nature of the first transition, and
> repeating the GOAWAY frame.
>
> Roberto would prefer that we create a new state in the flow:
>
>   accepting requests
>     --[DRAINING]-> about to stop accepting requests
>     --[GOAWAY]-> not accepting requests
>     --[TCP FIN]-> closed
>
> I note that this DRAINING frame looks similar in function to ALTSVC.
>
> If this summary is accurate, and in the absence of new information, I
> think that the best plan for this is to close at the interim, once we
> have more feedback from implementations.
>
>
Received on Thursday, 8 May 2014 22:01:38 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 17:14:30 UTC