[asterisk-dev] Idle Timers and Keep-Alives

[marek cervenka]

+1


Dne 24/01/2018 v 21:50 George Joseph napsal(a):
> So here's a proposal...
>
> We remove BOTH pjproject and asterisk keep-alives.
> We add the following parameters to pjsip transport:
>
> tcp_keepalives = <boolean> ; turn on or off tcp keepalives
> ; If tcp_keepalives = yes, the following parameters can be used to 
> override the default kernel
> ; settings in /proc/sys/net/ipv4/tcp_keepalive_(time,intvl,probes)
> tcp_keepalive_time = <seconds> ; The number of seconds a connection 
> can be idle before the first keepalive is sent.
> tcp_keepalive_interval = <seconds> ; Interval between keepalive probes.
> tcp_keepalive_probes = <number> ; Number of unacknowledged probes 
> before a failure is reported.
>
> To preserve backward compatibility, the current keep_alive_interval 
> setting in pjsip.conf/global
> would turn tcp keepalives on for tcp and tls transports with both 
> tcp_keepalive_time and
> tcp_keepalive_interval set to keep_alive_interval and 
> tcp_keepalive_probes set to 2.
>
> One advantage of this is that wireshark captures will clearly show 
> these as tcp keepalives even on
> a tls connection.   Another advantage is that we eliminate the 
> competing keepalive mechanisms
> with their threading and locking baggage.
>
> No code changes to pjproject would be required for this change.  We 
> can turn off their keepalives
> in our config_site.h file.
>
>
>
>
> On Fri, Jan 5, 2018 at 9:07 AM, Ross Beer <ross.beer at outlook.com 
> <mailto:ross.beer at outlook.com>> wrote:
>
>     Could the operating system manage this also, for example with the
>     following:
>
>     sysctl.conf
>
>     net.ipv4.tcp_fin_timeout = 60
>     net.ipv4.tcp_retries1 = 3
>     net.ipv4.tcp_syn_retries = 5
>
>     # Keep TCP connections alive
>     net.ipv4.tcp_keepalive_time = 300
>     net.ipv4.tcp_keepalive_intvl = 60
>     net.ipv4.tcp_keepalive_probes = 20
>
>     From a chan_pjsip point of view, it would receive notification
>     that the underlying connection has closed.
>     ------------------------------------------------------------------------
>     *From:* asterisk-dev-bounces at lists.digium.com
>     <mailto:asterisk-dev-bounces at lists.digium.com>
>     <asterisk-dev-bounces at lists.digium.com
>     <mailto:asterisk-dev-bounces at lists.digium.com>> on behalf of
>     Alexander Traud <pabstraud at compuserve.com
>     <mailto:pabstraud at compuserve.com>>
>     *Sent:* 05 January 2018 15:44
>     *To:* Asterisk Developers Mailing List
>     *Subject:* Re: [asterisk-dev] Idle Timers and Keep-Alives
>     > Do we even WANT an idle timer?
>
>     I posted my concerns already in <http://gerrit.asterisk.org/6807
>     <http://gerrit.asterisk.org/6807>>: I have a device which crashes
>     when it receives such a keepalive. I could live with a timer when
>     Asterisk is not the registrar but registered somewhere else. But I
>     do not _need_ that either.
>
>
>
>     -- 
>     _____________________________________________________________________
>     -- Bandwidth and Colocation Provided by http://www.api-digital.com --
>
>     asterisk-dev mailing list
>     To UNSUBSCRIBE or update options visit:
>     http://lists.digium.com/mailman/listinfo/asterisk-dev
>     <http://lists.digium.com/mailman/listinfo/asterisk-dev>
>
>     --
>     _____________________________________________________________________
>     -- Bandwidth and Colocation Provided by http://www.api-digital.com --
>
>     asterisk-dev mailing list
>     To UNSUBSCRIBE or update options visit:
>     http://lists.digium.com/mailman/listinfo/asterisk-dev
>     <http://lists.digium.com/mailman/listinfo/asterisk-dev>
>
>
>
>
> -- 
> George Joseph
> Digium, Inc. | Software Developer
> 445 Jan Davis Drive NW - Huntsville, AL 35806 - US
> Check us out at: www.digium.com <http://www.digium.com/> & 
> www.asterisk.org <http://www.asterisk.org/>
>
>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.digium.com/pipermail/asterisk-dev/attachments/20180126/48dd8e96/attachment.html>