All RTP Voice Traffic going through server!?!?!?!?!

Discussion in '3CX Phone System - General' started by greg@summitrad, May 16, 2017.

Thread Status:
Not open for further replies.
  1. greg@summitrad

    Joined:
    Aug 15, 2014
    Messages:
    27
    Likes Received:
    0
    We have been having horrendous audio quality issues and as such, we’ve been in troubleshooting mode. We're running V15 with the latest update, and our phones are almost entirely Yealink T46G (we have a few T46S now as well) We had some hardware issues at our Main office, and moved our phone system to our bigger VMWare Cluster at the Data Center. We are connected by several links (50 and 100 meg links) between our main office and where our Datacenter is. We are only at about 20-30% bandwidth use on any link. We record all the phone calls on virtually every extension. (We’ve disabled all but 10 recording extensions for troubleshooting). We have some new hardware here now, and we’re planning to bring the phone system back in house.

    However, In the process of troubleshooting, we discovered a very , very disturbing issue. EVERY last drop of voice traffic is being funneled through the phone system. EVEN IF recording is disabled, even if there is no checkbox in the “PBX Delivers Audio” box ALL audio is transmitted via the 3CX Server. No boxes selected in troubleshooting at all and Everything still flows through the PBX.

    90% of the phones are in the same building as our Patton PRI gateway. We have Patton FXS boxes for faxing, and all of that traffic is being funneled through the phone system as well. Only SIP should be going to the phone system, and all RTP traffic should be going between the phone and the PRI box. Even two extensions talking to each other (sitting next to each other on the same subnet in the same office) are going through the phone system.

    We’ve captured this with Wireshark mirroring our entire voice vlan. Sure enough, every last conversation has to go down to the data center and back up. So it’s coming into the PRI Up here in our main office, then it goes down to the phone system, then back up to the extension up here in our main office. We have some SIP trunks that terminate into the phone system in the data center. We’d expect those to be funneled through the PBX since it’s all terminated there. Having all calls bounce up and down the WAN is highly suspect.

    What is the reasoning for funneling all the voice traffic through the PBX eve if there is a SIP gateway and all our phones are talking SIP. Cisco, Altigen, Avaya work like they are supposed to and SIP is just for signaling. Endpoint to Endpoint talks RTP and bypasses the PBX for audio. Extension to Extension calls, not being recorded – there is no possible reason I can think of that with a VoIP phone system to have those calls go all the way back to the phone system and back out to the phones. It’s Crazy.

    If anyone has an explanation, I would appreciate it. It’s counterintuitive and defeats the purpose of deploying the SIP/RTP combination.

    Greg
     
  2. leejor

    leejor Well-Known Member

    Joined:
    Jan 22, 2008
    Messages:
    10,585
    Likes Received:
    252
    From your description it's still not clear if the 3CX server is on the same LAN as the majority of sets, and the PRI gateway. If not, are you using an SBC at the remote location? Any call that is recorded will have to route back to the server for recording to take place, as you seem to be aware. You will have to explain in a bit more detain, how your network is set up.

    Could it be a Codec, transcoding issue?
     
  3. greg@summitrad

    Joined:
    Aug 15, 2014
    Messages:
    27
    Likes Received:
    0
    As I stated above, for troubleshooting, we've disabled recording on nearly every extension to eliminate that cause for traffic to flow through the phone system. Our Handsets and PRI gateway are still in this same building, but different subnets. Our PBX was on a different subnet, even when it was in the same building with the PRI gateway and handsets. So it's almost identical now, Handsets are in a VoIP subnet, PRI gateway (and FXS boxes) are in another subnet, and the Phone system exists in another subnet. (We did move the PBX Subnet to our datacenter when we moved the phone system down there. We wanted to keep it all the same for licensing purposes). Same machine, same mac, same address, same activation. In the course of troubleshooting, we did build an entirely NEW phone system and activate it as such. Same horrendous audio.

    We are not using SBCs.

    This voice degradation was not an issue prior to moving the phone system to our datacenter.

    We have all Cisco switching with QOS defined, and the switches are correctly marking the traffic. All switches and NICs on the VMWare cluster in our data center are 10Gbps switches. Just for kickes, we did isolate all voice traffic onto a separate 1 Gbps network as a troubleshooting step. It STILL is being sent through the PBX.

    We have bandwidth to spare, and Call admission control is not a problem. We have 23 lines, and even running wide out on the largest codec, we can't saturate our bandwidth. (Since 3CX saw fit to remove the reports telling us how many lines we have in use at any hour, I can't tell you what our max lines in use is, but it is not 23).

    We do no video and no chat via 3CX for windows. We do not use any softphone functionality, nor do we use any mobile functionality via Android or IOS. Our server is Windows 2012 R2. AV had even been disabled (and the Firewall) to rule it out.

    We've tried each codec Available in 3CX (PCMU (G 711U), PCMA (G 711a), G 722, and G729) We recieved the same results on all codecs. The voice traffic is funneled through the PBX.

    In Comparison, our SIP Provider offers a cloud phone system that we use as an emergency backup. We've set that up on several phones, all via SIP/RTP of course, and added extensions to to the phones as a different "account" in the yealinks. When we do a packet trace on that, We do see our SIP traffic for voice going out to the provider, but I would expect that because they are the termination endpoint for those SIP trunks. If I call from Extension to Extension (as I have with 3CX.. on a phone on the same subnet sitting 1 desk away from me), I see the SIP signaling to set the call up, but the phones are talking RTP directly to each other.
     
Thread Status:
Not open for further replies.