you can use send_reply() to avoid any test in the config.

Thanks for the tip Daniel, will give it a go!

So think that kamailio itself is the branch of the incoming request, and can create many outgoing branches. If it generates a final reply, it does it to the incoming branch. It is not forwarding (sending back) a reply from an outgoing branch.

Ok, when you put it that way it actually makes more sense - thanks for a great example!

I still have one question left about my current scenario - why does Kamailio retransmit final non-2xx response after receiving an ACK if the response is valid and, as far as I can tell, so is the ACK?

Just to provide more context: 1) Ingress INVITE vs 504 response from Kamailio

2) 504 response from Kamailio vs Caller ACK

Regards, Ivan

On 11 Dec 2020, at 14:06, Daniel-Constantin Mierla miconda@gmail.com wrote:

Hello,

actually you are wrong that you think it is a wrong to tag. The entity that generates the final reply sets the to-tag (e.g.,to be able to match ack). Even here where you talk about the serial forking, you are going to face the case of different to-tag values for each branch sent out and replied. But also think about parallel forking, that timed out with many active branches, which to-tag to use?!?

So think that kamailio itself is the branch of the incoming request, and can create many outgoing branches. If it generates a final reply, it does it to the incoming branch. It is not forwarding (sending back) a reply from an outgoing branch.

Obviously, being open source, one can code a specific behaviour for some specific needs, but is not SIP specs compliant. If the upstream cannot deal with different to-tags, it won't really work with serial/parallel forking anyhow.

As a side remark, seeing the comments, if route(ROUTE_ELEMENT_FINAL_FAILURE) is used only in failure_route, then transaction is always defined there, because failure_route is executed only in the transaction context. Anyhow, if you are using in many places, not knowing if transaction was created, you can use send_reply() to avoid any test in the config.

Cheers, Daniel

On 11.12.20 11:04, Ivan Ribakov wrote:

...

Hi all,

I’m trying to implement following routing behaviour:

• ingress INVITE is sequentially forked to 1 or more endpoints
• under certain conditions sequential forking may be stopped and pre-defined error returned to the caller

Here’s is part of the config responsible for handling errors:

failure_route[ROUTE_ELEMENT_FINAL_FAILURE_HELPER] { route(ROUTE_ELEMENT_FINAL_FAILURE); }

route[ROUTE_ELEMENT_FINAL_FAILURE] { # We need to check if the error happens before the first egress, because T-state is not defined yet so we can not use t_reply if ($var(before_first_egress) == 0) { t_reply("$var(sipCode)", "$var(sipMessage)"); } else { sl_send_reply("$var(sipCode)", "$var(sipMessage)"); } }

Here is a simple scenario where delivering call to first endpoint times out after receiving 180 Ringing (using custom fr_inv_timer value) where:

• .33 is caller
• .208 is Kamailio
• .138 is callee

Notice that 504 is retransmitted by Kamailio. Here is a comparison of 180 and 504 returned by Kamailio:

As can be seen, 180 was proxied from the callee and has callee’s to-tag (tag=CALLEE1), however 504 that was generated by Kamailio using t_reply() has a different to-tag (presumably generated by Kamailio).

How can I tell Kamailio not to generate new to-tag but to use the last one that proxied back to the caller?

Thanks in advance for any help!

Regards, Ivan

---

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

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