TIPS Dual internet lines: Dropped calls

[LesD]

I have just installed a 2nd line for backup purposes and am using a Draytek 2850 in load balance mode. The purpose is to have a backup line but I might as well use both.

What then happened is that while outgoing calls via sip trunks were fine, incoming calls would ring, would be answered, and after 3-4 seconds the call would be dropped. A proper connection was being made as I could hear the other side and I think they heard me (not certain though).

I solved the problem by setting up a load balance policy to route all traffic from the PIAF box via a specific internet line and only use the 2nd one in the event of fall-over.

So problem solved, but I would like to understand what was going on. I would have thought that once the trunks register (they all work that way) the communication route would be fixed via a specific line. What was causing all the calls to be dropped?

 

[LesD]

Point taken but my understanding (very limited) was that the IP address known to Asterisk was required for connections from external SIP devices. As my trunks all register with the service provider, I would expect that connection to be treated as originating internally and therefore not a NAT issue.

Just come to mind is the fact that at an other location some time back I also had dual lines and never experienced this issue. That was on an older system using a different router so not directly comparable.

In case it is relevant, my system consists of

─SYSTEM INFORMATION───────────────────────────┐
│ │
│ PIAF Installed Version = 2.0.6.3 under *VIRTUALBOX* │
│ FreePBX Version = 2.10.1.9 │
│ Running Asterisk Version = 1.8.18.0 │
│ Asterisk Source Version = 1.8.18.0 │
│ Dahdi Source Version = 2.6.1+2.6.1 │
│ Libpri Source Version = 1.4.12 │
│ IP Address = 10.27.27.245 on eth0 │
│ Operating System = CentOS release 6.3 (Final) │
│ Kernel Version = 2.6.32-279.14.1.el6.i686 - 32 Bit

 

[atsak]

Almost certainly the load balance is jumping the RTP stream to a different outbound WAN link from the SIP signalling stream; or it is trying to move the traffic through both connections. Can't be done in SIP; must be one IP only.

 

[LesD]

I'm not trying to be difficult, but I would like to fully understand what is going on.

I still have several things I do not understand:

1. My old setup did not have this problem. so what is there different? It could well be that my old router did not load-balance properly or it was clever enough to recognise SIP traffic (unlikely).

2. Why is there no problem with outgoing calls?

3. Once an internet session (or whatever it is called) is established then the router will not move that session to an other line. As I am using trunks that are registered with the SIP service provider, I would expect all subsequent traffic to be restricted to the one line.

 

[atsak]

RTP streams are not necessarily delivered from the same IP or provider as the SIP trunk is registered to, so that could be one thing. The router may be load balancing by source IP, which is why you don't have problems with outbound (they come from the same IP).

The only way you could actually know for sure is to do a pcap on the data stream then analyze it in wireshark to see where the packets are going.

 

[LesD]

That makes sense. Thanks.