[Asterisk-Users] Re: failover (was Re: voicepulse)

Matt Lawson matt at 1control.com
Wed Jan 14 09:59:37 MST 2004


OK, so I answered my own question.  Turns out case #2 just goes to 
extension 2.

Still trying to figure out the optimum arrangement so I don't have an 
inordinate number of extensions.  Maybe like this:

1.  First outgoing try
2.  Second outgoing try
3.  Third ougoing try
4.  Play a message and/or hangup

102. Goto 2
203. Goto 3
304. Goto 4

>> But this is not to say _you_ can't built a reliable VOIP based
>> system.  Get _two_ providers and set up your dial plan in
>> extensions.conf to "fail over" if one service fails to
>> connect to dial via the next one and finally if both fail
>> use pstn. your users will see a system the "just works".
>
>
> Now there's an idea. 
> I'm playing with this now, but there's at least 1 case I'm having 
> trouble recognizing:
>
> The call connects but then drops due to "unauthorized."  It then only 
> goes to the "h" extension and I don't get a chance to try again.  Is 
> there anyway to detect this?
>
>
> I have to cover all of the following cases:
>
>
> 1.  VOIP IP address is not reachable.  Goes to extension n+101 (seems 
> to work as expected)
>
> 2. VOIP service answers but refuses with call with "unauthorized".  It 
> just goes to the "h" extension  Is there any watch to catch this 
> failure?  Perhaps put a timer on it and say if the call was less than 
> 5 seconds or something try the next one?
>
> Yes I am using a correct username and password and getting this today 
> (not from Voicepulse, from another provider).  But there's also a 
> moderate chance that during our systems' setup a name or password 
> could be misspelled so I need to cover this case.
>
> 3.  VOIP service connects but reports "all busy."  Well this one is 
> hard to test.  But I can make the Zap channel busy.  It goes to 
> extension n+101 as expected, so I'll have to assume that a busy VOIP 
> service does the same thing.
>
> I was trying to determine if the "t" or "h" extension would be useful 
> for these but I think not.  The timeout has to be set long enough for 
> someone to actually answer (20-60 sec or whatever).  The "h" is always 
> visited at the end of the call, whether it was sucessful or not.
>
> Any other cases, or suggestions how to handle case #2?
>
>





More information about the asterisk-users mailing list