The CUCM SRND states that the best way to implement a SIP trunk is to set the DMTF Preference on the Trunk in CUCM to "No Preference" and use dtmf-relay sip-kpml rtp-nte on the dial-peers pointing to the CUCM server(s). This allows CUCM to decide which method to use, depending if a phone supports out of band (e.g. KPML, SIP notify, etc.) or RFC2833 & thus avoid invoking an MTP to handle DTMF inter-working unless unavoidable.
Unfortunately this doesn't work with the old 7936 conference phones, you'll get no DTMF in calls via the SIP trunk, bug reference CSCuc80321 explains in detail. In short the workaround is to use DTMF Preference "RFC2833" on the Trunk, but this will always invoke an MTP. This in turn can cause a problem if you're using G729, as the MTPs running on the CUCM server(s) as part of the IP Voice Media Streaming App service only support G711. You'll need a router to be able to provide an MTP that can handle G729, as the example configuration snippet shows:
dspfarm profile 1 mtp
codec g729r8
codec g729ar8
maximum sessions software 60
associate application SCCP
Showing posts with label SIP trunk. Show all posts
Showing posts with label SIP trunk. Show all posts
Thursday, 13 August 2015
Friday, 23 March 2012
Kludging NAT for Communications Manager SIP Trunks
Ideally when using a SIP trunk to an external provider for PSTN connectivity with Communications Manager you'd use a Unified Border Element (UBE) in media flow-through mode to handle SIP to SIP inter-working & present a single IP address for NATing on your Internet router. But what happens when the customer is too cheap to pay for a UBE? Well if the SIP trunk requires registration or the SIP headers formatting in a way the CUCM doesn't like then you're out of luck! Otherwise you can kludge an IOS router into handling the NATing to a single IP address.
The problem stems from the fact that unless you enable "MTP required" Communications Manager handles a SIP trunk like an UBE in media flow-around mode - it handles protocol inter-working but the RTP audio streams pass directly between endpoints. This means that you have to NAT port 5060 to the CUCM server for the SIP traffic, but also NAT the phone's subnet too. Even if you do enable using an MTP, with multiple servers there could be multiple MTPs with different IP addresses.
The configuration below works assuming your IP address for NAT is 1.2.3.4, the CUCM server is 10.1.1.2, the phones reside in the 10.1.1.0/24 network & the SIP trunk is 4.3.2.1:
ip inspect name FIREWALL tcp
ip inspect name FIREWALL udp
ip inspect name FIREWALL sip
!
interface FastEthernet0
description Outside interface
ip address 1.2.3.1 255.255.255.248
ip access-group OUTSIDE_IN in
ip nat outside
ip inspect FIREWALL in
ip inspect FIREWALL out
interface FastEthernet1
description Inside interface
ip address 192.168.0.1 255.255.255.0
ip nat inside
!
ip access-list extended OUTSIDE_IN
permit tcp host 4.3.2.1 host 1.2.3.4 eq 5060
permit udp host 4.3.2.1 host 1.2.3.4 eq 5060
deny ip any any log
access-list 2 remark Voice subnet
access-list 2 permit 10.1.1.0 0.0.0.255
!
ip nat pool phones 1.2.3.4 1.2.3.4 netmask 255.255.255.252
ip nat inside source list 2 pool phones overload
ip nat inside source static udp 10.1.1.2 5060 1.2.3.4 5060 extendable
This kludge works as we've done a static port NAT to the CUCM server & allowed it in the access-list, this allows call control messages in & out. The SIP inspect punches holes in the access-list to allow the RTP inbound (outbound is allowed anyway). The SIP NAT ALG (application layer gateway) fixes the SIP headers to have the right IP addresses & port numbers. You can confirm this via show ip nat translations which shows the phone's being NATed as well as the CUCM server:
Pro Inside global Inside local Outside local Outside global
udp 1.2.3.4:5060 10.1.1.2:5060 4.3.2.1:5060 4.3.2.1:5060
udp 1.2.3.4:5060 10.1.1.2:5060 --- ---
udp 1.2.3.4:32366 10.1.1.21:32366 4.3.2.1:30118 4.3.2.1:30118
udp 1.2.3.4:32367 10.1.1.21:32367 4.3.2.1:30119 4.3.2.1:30119
udp 1.2.3.4:26642 10.1.1.24:26642 4.3.2.1:21068 4.3.2.1:21068
udp 1.2.3.4:26643 10.1.1.24:26643 4.3.2.1:21069 4.3.2.1:21069
udp 1.2.3.4:19462 10.1.1.26:19462 4.3.2.1:31900 4.3.2.1:31900
udp 1.2.3.4:19463 10.1.1.26:19463 4.3.2.1:31901 4.3.2.1:31901
udp 1.2.3.4:29940 10.1.1.26:29940 4.3.2.1:20970 4.3.2.1:20970
udp 1.2.3.4:29941 10.1.1.26:29941 4.3.2.1:20971 4.3.2.1:20971
udp 1.2.3.4:25306 10.1.1.28:25306 4.3.2.1:25398 4.3.2.1:25398
udp 1.2.3.4:25307 10.1.1.28:25307 4.3.2.1:25399 4.3.2.1:25399
The problem stems from the fact that unless you enable "MTP required" Communications Manager handles a SIP trunk like an UBE in media flow-around mode - it handles protocol inter-working but the RTP audio streams pass directly between endpoints. This means that you have to NAT port 5060 to the CUCM server for the SIP traffic, but also NAT the phone's subnet too. Even if you do enable using an MTP, with multiple servers there could be multiple MTPs with different IP addresses.
The configuration below works assuming your IP address for NAT is 1.2.3.4, the CUCM server is 10.1.1.2, the phones reside in the 10.1.1.0/24 network & the SIP trunk is 4.3.2.1:
ip inspect name FIREWALL tcp
ip inspect name FIREWALL udp
ip inspect name FIREWALL sip
!
interface FastEthernet0
description Outside interface
ip address 1.2.3.1 255.255.255.248
ip access-group OUTSIDE_IN in
ip nat outside
ip inspect FIREWALL in
ip inspect FIREWALL out
interface FastEthernet1
description Inside interface
ip address 192.168.0.1 255.255.255.0
ip nat inside
!
ip access-list extended OUTSIDE_IN
permit tcp host 4.3.2.1 host 1.2.3.4 eq 5060
permit udp host 4.3.2.1 host 1.2.3.4 eq 5060
deny ip any any log
access-list 2 remark Voice subnet
access-list 2 permit 10.1.1.0 0.0.0.255
!
ip nat pool phones 1.2.3.4 1.2.3.4 netmask 255.255.255.252
ip nat inside source list 2 pool phones overload
ip nat inside source static udp 10.1.1.2 5060 1.2.3.4 5060 extendable
This kludge works as we've done a static port NAT to the CUCM server & allowed it in the access-list, this allows call control messages in & out. The SIP inspect punches holes in the access-list to allow the RTP inbound (outbound is allowed anyway). The SIP NAT ALG (application layer gateway) fixes the SIP headers to have the right IP addresses & port numbers. You can confirm this via show ip nat translations which shows the phone's being NATed as well as the CUCM server:
Pro Inside global Inside local Outside local Outside global
udp 1.2.3.4:5060 10.1.1.2:5060 4.3.2.1:5060 4.3.2.1:5060
udp 1.2.3.4:5060 10.1.1.2:5060 --- ---
udp 1.2.3.4:32366 10.1.1.21:32366 4.3.2.1:30118 4.3.2.1:30118
udp 1.2.3.4:32367 10.1.1.21:32367 4.3.2.1:30119 4.3.2.1:30119
udp 1.2.3.4:26642 10.1.1.24:26642 4.3.2.1:21068 4.3.2.1:21068
udp 1.2.3.4:26643 10.1.1.24:26643 4.3.2.1:21069 4.3.2.1:21069
udp 1.2.3.4:19462 10.1.1.26:19462 4.3.2.1:31900 4.3.2.1:31900
udp 1.2.3.4:19463 10.1.1.26:19463 4.3.2.1:31901 4.3.2.1:31901
udp 1.2.3.4:29940 10.1.1.26:29940 4.3.2.1:20970 4.3.2.1:20970
udp 1.2.3.4:29941 10.1.1.26:29941 4.3.2.1:20971 4.3.2.1:20971
udp 1.2.3.4:25306 10.1.1.28:25306 4.3.2.1:25398 4.3.2.1:25398
udp 1.2.3.4:25307 10.1.1.28:25307 4.3.2.1:25399 4.3.2.1:25399
Subscribe to:
Posts (Atom)