[Asterisk-Dev] Ring requested on channel already in use

[Matt Fredrickson]

On Tue, Sep 06, 2005 at 03:56:11PM -0400, alan wrote:
> A problem was recently posted on the Asterisk-Users mailing list, and it
> went unresolved. Now that it's plaguing our production system as well, I
> need to look into it further.
> 
> I am fairly sure this is a bug and not a misconfiguration issue. I will
> be happy to provide much more additional information, after I'm sure
> it's a bug and not a -users issue :)
> 
> Summary:
> 
> Occasionally, the same PRI channel receives two "Channel ... got hangup"
> notifications in sequence, with no intervening activity on the channel.
> A subsequent call coming in to the same span and channel receives the
> "warning:"
> Ring requested on channel 0/3 already in use on span 1. Hanging up owner.
> 
> When this happens, Asterisk enters an inconsistent state, which I have
> only been able to correct by restarting.
> 
> Some symptoms I've seen when it's in this state are: previously active
> "asterisk -r" sessions fail to exectue commands; most/all extensions
> fail when dialed by local phones; and of course incoming calls on the
> PRI fail (with the Ring Requested error).
> 
> A few more details:
> 
> I dug into chan_zap.c, and found 2 places which generate the exact same
> "Channel 0/3, span 1 got hangup" message, both in the huge function
> pri_dchannel. One was inside "case PRI_EVENT_HANGUP" and one was inside
> "case PRI_EVENT_HANGUP_REQ".
> 
> I recompiled with slightly different error messages, so I could see what
> was happening in this case. In the one failure case I've had since that
> change, the first "Channel got hangup" message was generated by the
> PRI_EVENT_HANGUP_REQ case, and the second message was generated by
> PRI_EVENT_HANGUP.
> 
> "set debug 4" and "pri intense debug" logs are available if anyone knows
> more about parsing the intense PRI debugging than I do.
> 
> 
> ** System specs:
> Asterisk 1.0.8
> libpri 1.0.8
> zaptel 1.0.8
> fedora core 3
> 1x Digium Wildcard TE110P, single T1/E1 card
>   with a T1/PRI to the PSTN (US)

Good report, lots of information.  See if you can reproduce it in CVS-HEAD
(Asterisk, libpri, zaptel)

-- 
Matthew Fredrickson