Hello,
well, the ACK is a request without a response, practically it is a stateless request, no sip transaction can be created for it because there is no response to wait for it. That's by design from SIP specs.
In other words, it is no way to know if the target received or not the ACK. If the ACK is not received, the target is supposed to retransmit the 200ok, to force retransmission of the ACK.
In the case of tcp/tls, one can leverage transport layer to know that it could not be transmitted, but for udp is no way, and even more, in sip transport layer is decoupled from sip singling layer (ie., same connection can carry traffic for many users, many transactions, etc...). Also, for tcp/tls, with asynchronous sending, the feedback of not able to send is not immediate. But can be coded somehow, it's about open source after all ...
Moreover, ACK does not belong to INVITE transaction and can have a different path than the INVITE, being a matter of record-route headers.
So lack (or limitations) of DNS failover for ACK comes from the above.
You can either consider this a corner case, if the target servers are supposed to run always and in case of one becoming unavailable for long time, the dns is updated accordingly, or, if you know that ACK has to be sent to the address where 200ok came from, then you can store that in htable and use it for sending out the ACK (but again, this may not be the case always according to the specs, but can be in your deployment).
Cheers, Daniel
On 23.02.22 15:59, Julien Klingenmeyer wrote:
Hello,
I use DNS failover feature for routing some calls, and I wonder about its limitations.
I noticed that INVITE and ReINVITE requests are correctly routed based on SRV priority/weight in case of failure, but ACK requests do not use them.
Let’s say I have Kamailio A relaying requests to a pool of Kamailios B1 and B2 in TLS.
DNS records are the below ones.
kamailiob.net. 60 NAPTR 20 100 S SIPS+D2T _sips._tcp.kamailiob.net.
_sips._tcp.kamailiob.net. 60 SRV 1 10 5061 kamailiob-1.net.
_sips._tcp.kamailiob.net. 60 SRV 2 10 5061 kamailiob-2.net.
If B1 is down, I expect requests being relayed to B2 (because of priority 2 in the SRV record).
- For initial INVITEs: R-URI is sip:kamailiob.net o If B1 is down, the request is retried to B2 => OK
- For in-dialog requests: Route header in incoming request is sip:kamailiob.net so it is set as next destination with /loose/_/route/ function o If B1 is down, the request is retried to B2 for ReINVITE => OK o But once the ACK comes in, with the same Route header (sip:kamailiob.net), Kamailio tries to send it to B1 only (no retries to B2, and even no retransmissions to B1 either). o When B1 is back UP again a few seconds later, Kamailio does not try to relay the ACK again. The ACK request is attempted to be retransmitted only once to B1. It fails and no more retries after that.
Is it the expected behavior? Or is it something misconfigured on Kamailio A? Or is it a TLS connection issue?
Here is the DNS configuration on Kamailio A (TM module is enabled):
dns_try_naptr=yes
use_dns_failover=yes
use_dns_cache=yes
dns_srv_lb=yes
enable_tls=yes
tcp_max_connections=4096
tls_max_connections=4096
tcp_reuse_port=yes
tcp_connect_timeout=10
tcp_send_timeout=10
tcp_connection_lifetime=3600
I had a look into the DNS tutorial (https://github.com/kamailio/kamailio/blob/master/doc/tutorials/dns.txt) without finding any hint about this.
If anyone has already played with SIP DNS failover in Kamailio, your help would be appreciated, thanks!
Julien
Kamailio - Users Mailing List - Non Commercial Discussions
- sr-users@lists.kamailio.org
Important: keep the mailing list in the recipients, do not reply only to the sender! Edit mailing list options or unsubscribe: