[asterisk-dev] Asterisk + DAHDI failure-resistance (non)

[Tzafrir Cohen]

On Sat, Jul 06, 2013 at 03:07:41PM +0300, Oron Peled wrote:
> Hi,
> 
> _Today's Asterisk+DAHDI state:_
>      *  If a single channel/span breaks...
>      *  On the next asterisk restart -- no DAHDI channels at all!
>      *  Is it acceptable in 2013?
> 
> This could have been fixed for years.
> 
> _What happened on the DAHDI side?_
>      *  In 2008[1] this problem was presented, with proposed solution.
>      *  Few months later we had a prototype[2] of the DAHDI side.
>      *  Tried to revive interest in 2010[3].
>      *  In 2011 Shaun Raffle started a big revamp of DAHDI in different direction. 

Err... Ruffell :-).

> This resulted in the pinned-spans[4] scheme which was part of DAHDI-*2.6.0*.
>      *  Since Nov-2011, it also included the required fixes for DAHDI-tools (see the 
> dahdi_cfg '-S<span>' and '-C<chan_list>' options).
>      *  Which means DAHDI is "ready" -- each channel/span may be handled 
> independently of the others.
> _What happened in Asterisk?_
>      *  Dynamic channel removal (DAHDI_EVENT_REMOVED)was added some 
> eons ago (in the Zaptel days).
>      *  It had a bug with D-Channels. Tzafrir uploaded a fix to review-board (726)
> [5] some 3 years ago -- it still lingers there instead of being applied to all 
There's a new version there (as of a month ago or so, that is). Unlike
previous versions, it will not change the deefault behaviour if you
don't need the new features. This turned out to be much simpler to do
than I thought earlier.

> supported asterisk versions.
>      *  Dynamic channel addition code was uploaded to review-board (1598)[6] in 
> Nov-2011 and keeps waiting there...
>      *  With these patches, Asterisk+DAHDI is fully dynamic, where we can 
> add/remove channels and spans at run-time without affecting the PBX.
>      *  But non of this is committed to trunk :-(
> 
> So with all the multi-year effort to solve this problem, we are still in square-one.
> 
> Releasing Asterisk 12 in this state is a major failure in my book.
> Are your customers happy when a single channel failure make the whole PBX 
> drop dead? If not, why don't you review those patches?

Another minor atvantage: it means that DAHDI devices can start after
Asterisk can. Thus there's no dependency of the Asterisk service on the
DAHDI "service". Which simplifies the service configurations.

In fact, this removes most of the need for any DAHDI "service" (all that
is left is a special script to unload all dahdi modules. As there's no
recursive rmmod).

-- 
               Tzafrir Cohen
icq#16849755              jabber:tzafrir.cohen at xorcom.com
+972-50-7952406           mailto:tzafrir.cohen at xorcom.com
http://www.xorcom.com