[asterisk-users] Re: FastAGI hangs up channel if server is not available

[Lee Jenkins]

Tony Mountifield wrote:
> In article <464C8977.4020501 at datatrakpos.com>,
> Lee Jenkins <lee at datatrakpos.com> wrote:
>> Running 1.2.14
>>
>> When I call a FastAGI script such as this script for an incoming call:
>>
>> [calldirect]
>> exten=>s,1,Answer()
>> exten=>s,2,AGI(agi://192.168.1.175/calldirect?check&${CALLERID(num)})
>> exten=>s,3,Goto(check_time,s,1)
>>
>> and the FastAGI server is not running (Asterisk gets "connection 
>> refused" TCP error), Asterisk just terminates the call like so:
>>
>> May 17 12:58:00 WARNING[10154]: res_agi.c:210 launch_netscript: Connect 
>> to 'agi://192.168.1.175/calldirect?check&NNNNNNNNNN' failed: Connection 
>> refused
>>    == Spawn extension (remove_caller, s, 2) exited non-zero on 
>> 'SIP/datatrak-0978b670'
>>
>> The text of the relevant wiki article 
>> (http://www.voip-info.org/wiki-Asterisk+FastAGI) infers that execution 
>> should roll through to the next executable line in the context where it 
>> was called in the case of an error in executing the FastAGI:
>>
>> "Asterisk 1.2
>> Under Asterisk 1.2, if a request to a FastAGI service failed for any 
>> reason, there was no way to determine this from the dialplan. The 
>> recommended action is to set a channel variable before calling the 
>> FastAGI, setting that variable to a known value within the FastAGI, and 
>> then checking that variable once the AGI has returned..."
>>
>> Is there a setting/var that I am missing?
> 
> No, the information is wrong. You need to make a small mod to res/res_agi.c
> in order to trap this case. Look for the second occurrence of
> LOCAL_USER_REMOVE, and add the lines just before it as show by this
> patch extract (ignore the line numbers; I have other mods too):
> 
> @@ -2065,6 +2106,9 @@
>                         close(fds[1]);
>                 if (efd > -1)
>                         close(efd);
> +       } else {
> +               pbx_builtin_setvar_helper(chan, "AGISTATUS", "NOTFOUND");
> +               res = 0;
>         }
>         LOCAL_USER_REMOVE(u);
>         return res;
> 
> With this change, a refused connection will not hang up the channel, but
> instead will set the AGISTATUS variable to "NOTFOUND".
> 
> Cheers
> Tony

Nice.  Thanks, Tony.

Has this patched been accepted/committed in subsequent releases such as 
a 1.2.17?

Thanks again,

Lee


-- 

Warm Regards,

Lee