[asterisk-dev] Idle Timers and Keep-Alives

[Ludovic Gasc]

2018-01-26 12:39 GMT+01:00 Alexander Traud <pabstraud at compuserve.com>:

> > You'd be OK with keeping an idle timeout it as long as you can turn
> > it off at runtime.
>
> Yes, I do not need it.
>
> I do not use keep-alive for SIP-over-TCP (or TLS), neither
> - within the TCP layer (via its own mechanism), nor
> - on the TCP layer (via CRLN).
> Even short SIP-Register must be used carefully as explained by the author
> of Zoiper: <http://vimeo.com/82094736> and <http://vimeo.com/109598906>.
> Although the videos are 2 x 45 minutes long, I recommend them, to
> understand the frustrations of us mobile VoIPers.
>

>From my experience, tcp keepalive is complementary with OPTIONS, because
some routers can close silently a TCP connection without data, even if the
tcp keep-alive is enabled.
I have found an issue in Chromium about that:
https://bugs.chromium.org/p/chromium/issues/detail?id=27400
Moreover, in this article:
https://blog.stephencleary.com/2009/05/detection-of-half-open-dropped.html
The paragraph "Manipulate the TCP/IP keepalive packet settings" tries to
explain the advantages and problems.

As explained on: http://blogs.asterisk.org/2018/01/27/wanted-dead-or-alive/
TCP keep-alive should help to decrease the frequency of OPTIONS, but it
needs to be measured on a production environment to be sure.
When the new release of Asterisk will be ready with these changes, we will
do some tests.


> > At what level of granularity would it need to be?
> > System, transport, endpoint?
>
> System, in my case.
>

Theoretically, endpoint level is the best in term of customization :-)
Nevertheless, in this specific use case, I think we will also use system
level: We have also permanent connections with WebSockets not related with
Asterisk from our customers, we will put a value that we will work on any
TCP connection between them and us.

Have a nice week.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.digium.com/pipermail/asterisk-dev/attachments/20180128/cdcd02e9/attachment.html>