Summary:ASTERISK-03136: DUNDI replies always over 2000ms
Reporter:Marc C (mstorck)Labels:
Date Opened:2004-12-29 18:47:57.000-0600Date Closed:2011-06-07 14:10:48
Versions:Frequency of
Description:On my server DUNDi can do 'alive' requests to any of my DUNDi peers, but it doesn't answer such 'alive' request in a timely manner (under 2000ms). The problem appears randomly, sometimes my peers get replies from the server.
The other services like SIP, IAX, etc etc work wonderful. The server load is 0.00 0.00 0.00 most of the time. ICMP pings to my peers are all under 150ms, most peers have times around 20ms.


Have a look at:
Comments:By: Mark Spencer (markster) 2004-12-29 18:48:59.000-0600

This is a configuration issue, not a bug.  You have dundi queries for contexts which are in the contexts you are querying thus creating a loop.

By: Marc C (mstorck) 2004-12-29 19:25:06.000-0600

i spent hours because of this in the chat on irc.freenode.net #dundi and #asterisk, I googled, yahoo'ed, I mailed the asterisk's mailing list and digium support, nobody had a clue what it is. So if all those people don't have clue what it could be... so if it's not a bug, what is it?

By: Marc C (mstorck) 2004-12-29 19:53:08.000-0600

just to bring some more light into this, i do not believe that there is a loop:


exten => _352800XXXXX,1,SetAccount(0.00)
exten => _352800XXXXX,2,Dial(ZAP/g1/${EXTEN:3})


; default extensions
exten => h,1,Hangup
exten => s,1,Hangup
exten => i,1,nocdr()
exten => i,2,Hangup
exten => t,1,nocdr()
exten => t,2,Hangup


dundi-test => DUNDI-E164,400,SIP,${NUMBER}@pri1.gateway.lu,nounsolicited,nocomunsolicit

The dundi switch is in the [DUNDI] context, which gets included in 2 other
contexts [context1] and [context2] where context1!=context2!=DUNDI-E164

This is the dialplan as seen by asterisk:

pri1*CLI> show dialplan DUNDI-E164
[ Context 'DUNDI-E164' created by 'pbx_config' ]
 '_352800XXXXX' => 1. SetAccount(0.00)                           [pbx_config]
                   2. Dial(ZAP/g1/${EXTEN:3})                    [pbx_config]

-= 1 extensions (2 priorities) in 1 contexts. =-

So there is NO loop.

By: Brian West (bkw918) 2004-12-29 20:20:28.000-0600

Its a loop thats all it could be please check your peers to ensure you don't have a query loop or find someone on IRC in #dundi to help fix the problem.


By: Marc C (mstorck) 2004-12-29 20:27:29.000-0600

I do not have a dundi switch inside an exported context, so what other situation could cause that loop.

BTW, in case your IRC nick is bkw, I talked with you about 30 minutes on IRC #dundi about this problem (that's about 2-3 weeks ago), but you didn't have an idea what the problem could be.



By: twisted (twisted) 2004-12-29 20:28:47.000-0600

Not a blocking issue.

By: Mark Spencer (markster) 2004-12-29 20:32:59.000-0600

I mean I know this has probably occured to you already, but since this isn't something we're seeing in general, don't you admit that there is at least a significant liklihood of being a configuration issue?  have you looked at the debugs?

By: Marc C (mstorck) 2004-12-29 20:44:36.000-0600

I spent hours a day for weeks now on IRC, pasting my config over and over again... just right now ***** (from *****) is trying to help, he saw a complete copy of my dundi.conf and extensions.conf. He didn't find anything either. I tried asking on the mailing list, nobody responded.

edited on: 01-03-05 11:04

By: Mark Spencer (markster) 2004-12-29 20:58:00.000-0600

have you tried using the "dundi store history" to look and see what's really taking the time?

By: Marc C (mstorck) 2004-12-29 21:52:52.000-0600

"dundi store history" works fine, but how do i visualize that history, I googled for this, checked the wiki, asked in #asterisk,#dundi and #asterisk-dev. Some recommended "dundi show history" which does not work on my system.

By: nick (nick) 2004-12-29 23:48:48.000-0600

Is this with latest CVS?

By: Mark Spencer (markster) 2004-12-30 12:02:19.000-0600

It just gives you rich information when you perform a "dundi show peer <foo>" or "dundi show peers"

By: Mark Spencer (markster) 2005-01-06 21:42:57.000-0600

What did you find out, is this still a problem or should we close this out?

By: Mark Spencer (markster) 2005-01-12 23:37:16.000-0600

Okay, last try here, is this still a problem or is this ready to close?

By: Marc C (mstorck) 2005-01-13 06:36:45.000-0600

edited for privacy reasons:

*CLI> dundi show peers
EID    Host                Model      Order      AvgTime  Status
00:    66.   (S) Symmetric  Primary    995 ms   OK (153 ms)
00:    24.   (S) Symmetric  Primary    Unavail  UNREACHABLE
00:    208.  (S) Symmetric  Primary    941 ms   OK (133 ms)
00:    24.   (S) Symmetric  Primary    834 ms   OK (158 ms)
00:    217.  (S) Symmetric  Primary    Unavail  UNREACHABLE
00:    62.   (S) Symmetric  Primary    1477 ms  OK (17 ms)
00:    206.  (S) Symmetric  Secondary  254 ms   OK (95 ms)
00:    66.   (S) Symmetric  Primary    Unavail  UNREACHABLE
00:    66.   (S) Symmetric  Primary    710 ms   OK (139 ms)
00:    65.   (S) Symmetric  Primary    1535 ms  OK (91 ms)
00:    217.  (S) Symmetric  Primary    2440 ms  OK (17 ms)

By: Mark Spencer (markster) 2005-01-13 08:49:29.000-0600

Okay so this shows you where your problems are.  You can simply dundi show peer <foo> where <foo> is a peer that takes a long time to respond to you and see what the deal is.

By: Mark Spencer (markster) 2005-01-16 02:21:51.000-0600

One last request for a response to what we've said.  Given some peers respond in 254ms, I think that's a reasonably clear indication that this is not a bug but something related to your peering with the other hosts, but i have a small amount of patience remaining for further diagnostics to confirm.

To keep this open though, you're going to have to be more responsive than you've been.

By: Marc C (mstorck) 2005-01-16 06:21:19.000-0600

Please have a look at http://dundi-map.netmonks.ca/info.php?eid=00:10:dc:3a:7a:5e you will see that I peer with hosts most of you peer with as well... So why should the hosts you peer with as well answer much slower for me than for you?

By: Mark Spencer (markster) 2005-01-16 10:28:12.000-0600

What was the result of the debug I asked you to do on 1/13/2005?  Once you see there are delays, you can look at the history and see which queries took a long time to execute, and you can also do some debugs to watch what is going on.

By: Mark Spencer (markster) 2005-01-20 22:40:08.000-0600

Okay, maybe we should just talk on IRC, this bug report is obviously taking way too long and is increasingly unlikely to actually be a bug.