No audio second incoming call in queue

[lean-ict]

I have a weird problem. The scenario is as following:

All external incoming callers are connected to queue with multiple agents with Yealink SIP T46S phones.
The are multiple external incoming calls waiting in the queue.

• Agent 1 taking caller 1 in queue -> no problem
• Agent 2 taking caller 2 in queue -> no problem
• Agent 1 ends caller 1 in queue and immediately takes next caller
• Agent 1 taking caller 3 in queue -> no audio both ways

If agent 1 would wait a minute end then takes the next caller, there is no problem.
What could cause this problem? Either there is perfect audio or there is no audio at all...

Details of set-up:
3CX PBX
Hyper-V VM behind pfsense firewall in datacenter. All NAT rules set as described, SIP ALG disabled.
Static outbound rule set. 3CX Firewall test 100% success.
Phones
Yealink SIP-T46S, provisioned by PBX.

Tried following settings.

• Provisioning: Remote STUN with default SIP and RTP ports.
• Provisioning: Remote STUN with seperate SIP and RTP ports per phone.
• Enabling and disabling 'PBX delivers audio' option
• Using different codecs, was G729, then used G722 and G711 U-law
• Checking bandwith; is 50/5mb for only 10 phones, no other pc's connected to this DSL connection.
• Provisioning using local SBC

In the attachment is a PCAP of a failed call.
Hope the community can help us out!

Kind regards,

Martijn

 

[lean-ict]

While testing further, we notice slightly different behaviour in the PCAP logging when using SBC.
The second incoming phone call disconnects when agent tries to pick up.
Please see the attached PCAP log for the scenario with SBC, the other in the main thread is a PCAP log when provisioning using remote STUN.

 

[eddv123]

Does sound like a STUN issue, STUN can be a real issue with audio ports being blocked on the Firewall as it will allocate sporadic ports for RTP. For example in the trace I can see both RTP streams the 3CX side is using 9096 and local side 14004.

How many phones are using STUN on this remote site ? and also is this issue occurring on local calls external or both ? The PCAP itself is only showing the traffic between the PBX and 1 extension (not via provider or between 2 handsets).

 

[lean-ict]

There are currently 8 phones (SIP T42S) using STUN default ports.
As a test scenario (PCAP file 2) the 3 SIP T46S phones that receive the external callers from the queue are provisioned using SBC. The PCAP recording is done on one of the SIP T46S phones. It showing the answering of calls coming form the PBX out of the queue.
The first PCAP file all phones are provisioned with STUN.

 

[lean-ict]

We've tested to call the queue from internal extension. Same issue, the call is being disconnected when agent tries to pick up.

 

[lean-ict]

Have tried to disable STUN on Yealink phone that is using SBC. Same issue, a second call in queue is disconnecting.

 

[YiannisH_3CX]

Hello @lean-ict

Can you try flashing a phone with the supported firmware and see if you can still replicate the issue. As far as i can see you are using a Lydis firmware which is not tested by us. As a test flash a phone with the supported firmware and see if the issue persists. https://www.3cx.com/support/phone-firmwares/

 

[eddv123]

Additionally 3CX only really recommend the STUN method now for 1 or 2 handsets per site:
https://www.3cx.com/docs/manual/configuring-ip-phones/#h.ul2fzupi6t22/#h.rff40x4owe3y

You have a STUN checklist on this link here also.

 

[lean-ict]

Tried to flash all phones to the supported firmware versions. Now it seems to be dropping less calls. Next thing to do is provisioning all phones via SBC. Will try that today and let you know.

 

[lean-ict]

After testing and tweaking a bit I finally resolved the issue.

Firstly it turned out that the customer was not very specific on the 'no audio' problem description.
What really happened is that every 3rd or 4th call was being dropped.
You will not hear a 'disconnected'-tone like the old days, so it seemed a no audio problem.

After disabling the options 'support re-invite' and 'Support replaces' in the SIP Trunk, everything is working fine now. Tried to bombard the PBX with 6 simultaneous calls, simulating over 30 calls this way without a single call being dropped...

By default, this option was turned on in the trunk's template. But what does this do?
I've read several posts where it says it cannot be disabled...

 

[lean-ict]

Monday morning... The problem is still occurring at the clients site. How can this be???
What else can be done the test this? Is a PCAP on the PBX usefull?

 

[Lee Cramman]

Not sure what your problem is but it sounds like the site needs to route everything through an SBC.

 

[lean-ict]

I've used a SBC for all extensions on the clients site. On all extensions the SBC is used as outbound proxy.
I've now enabled the option 'Force invites to be send to IP of registrar' on the SIP trunk.
Now waiting to see if this helps...

I'm confused why the seems to occur at random. During my test I've been able to bombard the PBX with inbound calls, without a single one dropping...

 

[Lee Cramman]

I've used a SBC for all extensions on the clients site. On all extensions the SBC is used as outbound proxy.
I've now enabled the option 'Force invites to be send to IP of registrar' on the SIP trunk.
Now waiting to see if this helps...

I'm confused why the seems to occur at random. During my test I've been able to bombard the PBX with inbound calls, without a single one dropping...

The problem with STUN is that you're dependant on a properly configured and compliant router / firewall. I've seen quite a few small businesses running on less than ideal networking hardware that often doesn't handle opening / forwarding of ports very well.

Given how inexpensive an SBC is I like to just sidestep the issue and use one anywhere with >1 endpoint.