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

Re: Graceful shutdown #458

From: Jeff Pinner <jpinner@twitter.com>
Date: Thu, 8 May 2014 16:32:53 -0700
Message-ID: <CA+pLO_hQn6+JMFhE97VAOn3OhgNEe+OHZU68vc8912akpAMy-g@mail.gmail.com>
To: Martin Thomson <martin.thomson@gmail.com>
Cc: Daniel Sommermann <dcsommer@fb.com>, Roberto Peon <grmocg@gmail.com>, HTTP Working Group <ietf-http-wg@w3.org>
On Thu, May 8, 2014 at 4:26 PM, Martin Thomson <martin.thomson@gmail.com>wrote:

> On 8 May 2014 16:09, Jeff Pinner <jpinner@twitter.com> wrote:
> > We have a SPDY proof of concept running that will send two GOAWAY frames
> on
> > shutdown:
> >
> > GOAWAY(max stream id, NO_ERROR)
> > (send ping and wait for pong or timeout)
> > GOAWAY(last good stream id, *some code*)
> > TCP FIN
>
> I would have thought that you would want to let the requests you have
> received complete, or is that not the plan here?
>
>
We do, specifically we send GOAWAY with max stream id, and then for some
period of time until we send the second GOAWY, we continue to accept
streams. Then when we send the second GOAWAY, we send it with the last
stream id we have decided to process (which could have been received after
the first GOAWAY). Then we reject new streams with REFUSED and gracefully
close.



> Note that you don't send the last "good" stream, you send the last
> stream that you passed on for processing.
>
Received on Thursday, 8 May 2014 23:33:20 UTC

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