We don't see the issue when using the standard lookup() function, only when we start manually appending branches with the results from  reg_fetch_contact()

Thanks
Matthew

On Wed, Mar 17, 2021 at 7:18 AM Daniel-Constantin Mierla <miconda@gmail.com> wrote:

Hello,


On 11.03.21 10:55, Marrold wrote:
Hi Daniel,

I didn't spot those TCPOPs functions, I'll give them a try and let you know how I get on.

Do you have any idea why Kamailio is intermittently selecting the wrong connection using ID vs peer address?


have you tried and got that with lookup("location") or only with your script-based reg_fetch_contact()?

Cheers,
Daniel


Thanks for the suggestions.
Matthew

On Wed, Mar 10, 2021 at 7:27 AM Daniel-Constantin Mierla <miconda@gmail.com> wrote:

Hello,

a while ago I did some work to make possible to specify the outgoing tcp connection id, see:

  * https://www.kamailio.org/docs/modules/stable/modules/tcpops.html#tcpops.f.tcp_set_otcpid

And the next function after it.

However, the testing was minimal, maybe not verifying the entire chain with t_relay()/forward(). Even more, the specifying of the outbound tcp connection was supposed to be done internally by the lookup("location"), the functions from tcpops being added for more config flexibility, suitable for single branch forwarding or branch_route blocks.

However, in your seem to do manual processing with reg_fetch_contacts(), not rely on lookup location. You can test with master and use $sbranch(...) and corresponding functions from pv module, instead of setting the r-uri and append_branch().

Cheers,
Daniel

On 10.03.21 06:00, Marrold wrote:
Hi,

I've done a bit more digging and realised that $conid is read-only, and only available for an inbound connection - so I dont think it will achieve what I need.

I did a bit more troubleshooting and observed the differences in the debug log between two identical calls:

This example failed - the INVITEs went out to the incorrect endpoint / TCP connection. The "ulc_conid" from the location table for the TCP endpoint is 13:

Mar  9 10:30:20 proxy-01 /sbin/kamailio[27014]: ERROR: <script>: Adding to main branch. ua: Yealink SIP-T42G 29.83.0.120 ru: sip:442079460000@10.0.130.218:5060 du: sip:203.0.113.1:1046 ulc_conid: 0 flags: 192 ci: 162e3e84-4036-48d7-a1d1-79cc92b7cea4-2-1@sip-server-01
Mar  9 10:30:20 proxy-01 /sbin/kamailio[27014]: ERROR: <script>: Adding to child branch. ua: Yealink SIP-T58 58.85.0.5 ru: sip:442079460000@10.0.130.241:50130;transport=TCP du: sip:203.0.113.1:50130;transport=tcp ulc_conid: 13  flags: 192 ci: 162e3e84-4036-48d7-a1d1-79cc92b7cea4-2-1@sip-server-01

The debug log shows the wrong connection was found by id, which in this case was 2, but should have been 13:

Mar  9 10:30:21 proxy-01 /sbin/kamailio[27014]: INFO: <script>: ONSEND: rm: INVITE ru: sip:442079460000@10.0.130.218:5060 du: sip:203.0.113.1:1046 proto: udp src: 185.28.212.61:5060 dest: 203.0.113.1:1046 ci: 162e3e84-4036-48d7-a1d1-79cc92b7cea4-2-1@sip-server-01
Mar  9 10:30:21 proxy-01 /sbin/kamailio[27014]: INFO: <script>: ONSEND: rm: INVITE ru: sip:442079460000@10.0.130.218:5060 du: sip:203.0.113.1:1046 proto: tcp src: 185.28.212.61:5062 dest: 203.0.113.1:50130 ci: 162e3e84-4036-48d7-a1d1-79cc92b7cea4-2-1@sip-server-01
Mar  9 10:30:21 proxy-01 /sbin/kamailio[27014]: DEBUG: <core> [core/tcp_main.c:1651]: _tcpconn_find(): found connection by id: 2
Mar  9 10:30:21 proxy-01 /sbin/kamailio[27014]: DEBUG: <core> [core/tcp_main.c:2545]: tcpconn_send_put(): found fd in cache (5, 0x7fedc49d2448, 2)

This example worked - the INVITEs went out to the correct endpoints / TCP connections. The "ulc_conid" from the location table for the TCP endpoint is still 13:

Mar  9 10:42:30 proxy-01 /sbin/kamailio[27016]: ERROR: <script>: Adding to main branch. ua: Yealink SIP-T42G 29.83.0.120 ru: sip:442079460000@10.0.130.218:5060 du: sip:203.0.113.1:1046 ulc_conid: 0 flags: 192 ci: 95ba0691-bdfc-4293-b99d-d46946cb5180-2-1@sip-server-01
Mar  9 10:42:30 proxy-01 /sbin/kamailio[27016]: ERROR: <script>: Adding to child branch. ua: Yealink SIP-T58 58.85.0.5 ru: sip:442079460000@10.0.130.241:50130;transport=TCP du: sip:203.0.113.1:50130;transport=tcp ulc_conid: 13 flags: 192 ci: 95ba0691-bdfc-4293-b99d-d46946cb5180-2-1@sip-server-01

The debug log shows the correct connection was found by peer address and determined the correct connection 13:

Mar  9 10:42:31 proxy-01 /sbin/kamailio[27016]: INFO: <script>: ONSEND: rm: INVITE ru: sip:442079460000@10.0.130.218:5060 du: sip:203.0.113.1:1046 proto: udp src: 185.28.212.61:5060 dest: 203.0.113.1:1046 conid: <null> ci: 95ba0691-bdfc-4293-b99d-d46946cb5180-2-1@sip-server-01
Mar  9 10:42:31 proxy-01 /sbin/kamailio[27016]: INFO: <script>: ONSEND: rm: INVITE ru: sip:442079460000@10.0.130.218:5060 du: sip:203.0.113.1:1046 proto: tcp src: 185.28.212.61:5062 dest: 203.0.113.1:50130 conid: <null> ci: 95ba0691-bdfc-4293-b99d-d46946cb5180-2-1@sip-server-01
Mar  9 10:42:31 proxy-01 /sbin/kamailio[27016]: DEBUG: <core> [core/tcp_main.c:1670]: _tcpconn_find(): found connection by peer address (id: 13)
Mar  9 10:42:31 proxy-01 /sbin/kamailio[27016]: DEBUG: <core> [core/tcp_main.c:2548]: tcpconn_send_put(): tcp connection found (0x7fedc49ce0e8), acquiring fd

Additionally I checked the connection IDs and IPs / Ports before a failed and working call and they had not changed, it was the same TCP connection - so it doesn't appear to be some interaction with a connection closing or changing. We also don't see this issue using the standard lookup() function.

Does anyone know why Kamailio is intermittently switching between finding by peer address and id, and why it's using the wrong ID? 

Thanks again
Matthew


On Tue, Mar 9, 2021 at 8:56 AM Marrold <kamailio@marrold.co.uk> wrote:
Hi all,

I'm currently adding a feature to our Kamailio configuration to fork calls based on user agent.

To do so I'm getting the registered endpoints with reg_fetch_contacts() iterating through and matching on them, then using seturi() / append_branch() and setting the dst-uri and flags as required.

However, calls are intermittently going to the wrong TCP connection, despite the logs showing the correct destination IP and port in the ONSEND route.

Looking in the location table I can see each registration has a connection_id, and the debug log indicates it's finding the connection by ID:

 DEBUG: <core> [core/tcp_main.c:1651]: _tcpconn_find(): found connection by id: 3

Is there a way to set the conid per branch? Is it necessary?

We're using kamailio 5.3.3 on Debian 10.

Thanks
Matthew

_______________________________________________
Kamailio (SER) - Users Mailing List
sr-users@lists.kamailio.org
https://lists.kamailio.org/cgi-bin/mailman/listinfo/sr-users

-- 
Daniel-Constantin Mierla -- www.asipto.com
www.twitter.com/miconda -- www.linkedin.com/in/miconda
Funding: https://www.paypal.me/dcmierla
-- 
Daniel-Constantin Mierla -- www.asipto.com
www.twitter.com/miconda -- www.linkedin.com/in/miconda
Funding: https://www.paypal.me/dcmierla