asterisk users - May 2021 - pjsip transport and dynamic WAN IP address (not: NAT)

Hello all!

I'm actually wondering about how to achieve fast fail handling for the pjsip
transport if underlying WAN IP address changes.


Following scenario:
Asterisk runs on a device holding ppp0, which provides the interface for
outbound
registration to ISP trunks (transport acts as client) and therefore the WAN IP.

If WAN IP changes after restarting ppp0 (and ppp0 device completely disappeared 
for some time), it takes pretty long until the transport realizes, that the WAN
IP
changed.

I tried dnsmanager to always provide correct external IP address in SIP requests
like this

external_media_address     : external.mydom.org
external_signaling_address : external.mydom.org
local_net                  : 192.168.0.0/255.255.0.0

and I can see, that after IP address change, the SIP requests (Register e.g.) 
provide the correct WAN IP - but transport didn't realize the change on low
level.

NOTICE[4061] dnsmgr.c: dnssrv: host 'external.mydom.org' changed from
46.r.x.y:0
to 79.t.a.b:0

It takes quite some long time (~13 minutes), until pjsip / asterisk detects a 
timeout or a network unreachable.

tlsc0x7ffa800203a8 TLS transport destroyed with reason 120110: Connection timed
out

After this point has been reached, the transport is destroyed and a new
transport
is generated - from now on, things are working fine again.

I added a global logging rule in iptables at first place in OUTPUT to see any 
packages which are handled - but I couldn't see any - therefore I think
things are
breaking even earlier.


My question: is there any possibility to programmatically force the restart of a
given transport (I don't want to restart all transports)? This could be
easily
done in ppp up scripts. Or is there a configuration option to reduce the timeout
until restart of the individual transport? Maybe in pjsip (connection timeout 
detection)?

Another idea would be to restart the relevant transports based on dnsmgr
detecting
new external IP.



Thanks
Michael

On 01.05.21 at 07:13 Michael Maier wrote:
> 
> Hello all!
> 
> I'm actually wondering about how to achieve fast fail handling for the
pjsip
> transport if underlying WAN IP address changes.
Forgot to mention:

- I'm using TLS!
- pjsip tries every 92s to send the Registration until the timeout comes up. I 
don't know where the 92s are coming from. The tls_keep_alive_interval is 90s
- but
that's not a keep alive package. The "normal" configured retry
time if a
Registration is not answered, is 60s (which is working fine).




Thanks
Michael