On 25 August 2017 at 12:05, Daniel-Constantin Mierla <miconda@gmail.com> wrote:

The peer exposing the issue seems to be a pre-REF3261 implementation (no branch parameter in Via header). Can you also show the invite sent by the peer?

Cheers,
Daniel

On 23.08.17 16:17, George Diamantopoulos wrote:
A clarification:

The two exchanges-examples I included in the original message are not from the same peer. The issue is reproducible every time with the problematic peer (first example). I only included another exchange from a different peer (so it should read peer2 where I censored IP addressed) for comparison and to prove a point.

On 23 August 2017 at 17:13, George Diamantopoulos <georgediam@gmail.com> wrote:

Hello all,

I'm having a weird issue with Kamailio failing to properly process an ACK received to a 487 it sent, resulting in retransmissions of the 487. I assume it's because it can't match the ACK to the transaction, but I could be wrong.

I'm using a modified version of the default configuration, so ACKs should be handled correctly. I haven't editted the WITHINDLG route in any way that would affect this (or at least I think).

In addition, ACKs to 487 from other UAs are processed correctly, and these transactions are handled by the same routes in kamailio configuration as the problematic one, so I'm inclined to believe it's UA-specific?

Here's an example transaction of the failed kind (results in kamailio retransmitting the 487):

myself:5060 -> peer:5060
-------------------------
SIP/2.0 487 Request Terminated
Via: SIP/2.0/UDP peer:5060
From: <sip:user@peer>;tag=116B5368-24D8
To: <sip:tel@myself>;tag=as655f6372
Call-ID: 84DC69F2-873811E7-8A639B5A-3D9194E8@peer
CSeq: 101 INVITE
Server: modCOM v2 SIP Server
Allow: INVITE, ACK, CANCEL, OPTIONS, BYE, REFER, SUBSCRIBE, NOTIFY, INFO, PUBLISH, MESSAGE
Supported: replaces, timer
Content-Length: 0

peer:49590 -> myself:5060
-------------------------
ACK sip:tel@myself:5060 SIP/2.0
Via: SIP/2.0/UDP peer:5060
From: <sip:user@peer>;tag=116B5368-24D8
To: <sip:tel@myself>;tag=as655f6372
Date: Wed, 23 Aug 2017 12:50:47 GMT
Call-ID: 84DC69F2-873811E7-8A639B5A-3D9194E8@peer
Max-Forwards: 10
Content-Length: 0
CSeq: 101 ACK

And here's another similar transaction which is successful (no retransmissions):

myself:5060 -> peer:5060
------------------------
SIP/2.0 487 Request Terminated
Via: SIP/2.0/UDP peer:5060;branch=z9hG4bKjbmvq4009gthskk0a6s1.1
From: <sip:user@anonymous.invalid;user=phone>;tag=599D7495-9ACE9E3-0A324A05
To: <sip:tel@anonymous.invalid:5060;user=phone>;tag=as65375e5d
Call-ID: 599D7495-007A5832@fath3pcu238
CSeq: 1 INVITE
Server: modCOM v2 SIP Server
Allow: INVITE, ACK, CANCEL, OPTIONS, BYE, REFER, SUBSCRIBE, NOTIFY, INFO, PUBLISH, MESSAGE
Supported: replaces, timer
Content-Length: 0

peer:5060 -> myself:5060
------------------------
ACK sip:tel@myself:5060;user=phone SIP/2.0
Via: SIP/2.0/UDP peer:5060;branch=z9hG4bKjbmvq4009gthskk0a6s1.1
From: <sip:user@anonymous.invalid;user=phone>;tag=599D7495-9ACE9E3-0A324A05
To: <sip:tel@anonymous.invalid:5060;user=phone>;tag=as65375e5d

Call-ID: 599D7495-007A5832@fath3pcu238
Max-Forwards: 69
Content-Length: 0
CSeq: 1 ACK

I can't pinpoint anything wrong with the first exchange, other than the fact that for some reason, the "less than" (<) sign in the from and to URIs is escaped as &lt in homer's GUI (which also breaks CSS rendering in Firefox, I had to clear this code out). However, these escaping characters are not visible with sngrep when capturing traffic normally, and neither when doing a select in homer's database directly, so I guess it's a rendering bug in homer-ui and can be ignored (unless someone has reason to believe otherwise).

Now the relevant portion of the debug log is:

Aug 23 16:47:12 modcom-sbc-1 kamailio[9750]: {1 101 ACK 64AA4E6C-874011E7-9A729B5A-3D9194E8@peer} 7(9760) exec: *** cfgtrace:request_route=[WITHINDLG] c=[/etc/kamailio/kamailio.cfg] l=223 a=24 n=t_check_trans
Aug 23 16:47:12 modcom-sbc-1 kamailio[9750]: {1 101 ACK 64AA4E6C-874011E7-9A729B5A-3D9194E8@peer} 7(9760) DEBUG: tm [t_lookup.c:1001]: t_check_msg(): msg id=104 global id=103 T start=(nil)
Aug 23 16:47:12 modcom-sbc-1 kamailio[9750]: {1 101 ACK 64AA4E6C-874011E7-9A729B5A-3D9194E8@peer} 7(9760) DEBUG: tm [t_lookup.c:459]: t_lookup_request(): start searching: hash=54992, isACK=1
Aug 23 16:47:12 modcom-sbc-1 kamailio[9750]: {1 101 ACK 64AA4E6C-874011E7-9A729B5A-3D9194E8@peer} 7(9760) DEBUG: tm [t_lookup.c:494]: t_lookup_request(): proceeding to pre-RFC3261 transaction matching
Aug 23 16:47:12 modcom-sbc-1 kamailio[9750]: {1 101 ACK 64AA4E6C-874011E7-9A729B5A-3D9194E8@peer} 7(9760) DEBUG: tm [t_lookup.c:641]: t_lookup_request(): no transaction found
Aug 23 16:47:12 modcom-sbc-1 kamailio[9750]: {1 101 ACK 64AA4E6C-874011E7-9A729B5A-3D9194E8@peer} 7(9760) DEBUG: tm [t_lookup.c:1070]: t_check_msg(): msg id=104 global id=104 T end=(nil)
Aug 23 16:47:12 modcom-sbc-1 kamailio[9750]: {1 101 ACK 64AA4E6C-874011E7-9A729B5A-3D9194E8@peer} 7(9760) exec: *** cfgtrace:request_route=[WITHINDLG] c=[/etc/kamailio/kamailio.cfg] l=231 a=2 n=exit

It explicitly states that no transaction is found, after initiating pre-RFC3261 (why?) transaction matching. However, even pre-3261 matching should work, as the from and to headers as well as call-id in request and repy are the same.

Any input would be greatly appreciated, thanks!

George
_______________________________________________
Kamailio (SER) - Users Mailing List
sr-users@lists.kamailio.org
https://lists.kamailio.org/cgi-bin/mailman/listinfo/sr-users
-- Daniel-Constantin Mierla www.twitter.com/miconda -- www.linkedin.com/in/miconda Kamailio Advanced Training - www.asipto.com Kamailio World Conference - www.kamailioworld.com