CUCM: TLS 1.2 and Legacy Phones

Hello world!

Today’s post will be quick and dirty but hopefully useful.

In Cisco Call Manager (CUCM) version 11.5(1)SU3 support was introduced for Transport Layer Security (TLS) versions 1.1 and 1.2. Prior to this release 1.0 was the only supported version. With security being the driving factor in much of the IT world these days, there is a push to secure everything to the highest available level, that includes the Collaboration environment. I have many customers that want to take advantage of the higher TLS version levels and that is a good thing, but there are gotchas.

If you know anything about Cisco IP phone communications you know that several services; and specifically the Corporate and Personal directory service are pre-configured applets that run between the phone and CUCM. In the case of the directory service(s) they talk using 8443 which is a secure port and thus uses a certificate to communicate (along with the Trust Verification Service (TVS)). That certificate is directly encrypted with the help of TLS. Because support for the newer/higher TLS versions has only recently come into play there are several generations of IP phones that do not support anything above TLS 1.0 (or 1.1 in some cases). This list of legacy endpoints includes the 7900 series and Cisco’s previous “Cadillac” the 9900 series as well as others.

If you have legacy endpoints and change your TLS version to something above 1.0 you will notice that the directory services on those endpoints will fail with a “Host not found” error. What is actually happening involves a failed TLS handshake between the phone and the Trust Verification Service (TVS) on CUCM. Because the phone cannot communicate with the TVS using TLS 1.2 the handshake fails and the directory service cannot be accessed.

So what are your options?

  1. Replace your legacy phones, there is always a financial fix and this is it.
  2. You can manually create a directory service that talks HTTP and assign it to the phones. You could make this an Enterprise Subscription if you want everyone to use it. The URL that you should use is: http://YOURCUCMFQDNorIPHere:8080/ccmcip/xmldirectory.jsp. Since this is an HTTP service you can use the IP instead of the FQDN. This will allow the service to function whether name services are functional or not. This will require user training but may make the most sense depending on the environment.
  3. Revert to TLS 1.0. This sounds easy, but there are gotchas here too.

The gotchas of going back…

Switching from TLS 1.0 to TLS 1.1 or 1.2 (If you are going to change, go to 1.2) is a relatively straight forward process. You log into the platform CLI of your CUCM publisher and any subscribers and issue the following command set tls min-version 1.2. When you enter this command you will be asked to confirm and once confirmed (with yes) the system will restart. This happens immediately and should only be done during a maintenance window.

Switching back from TLS 1.2 to 1.0 follows the same steps with the same command and the same reboot process. However, when you switch back from a higher TLS version to a lower TLS version, anything that was encrypted using that higher TLS version becomes invalid. This includes certificates (which automatically regenerate) and also the application UI password store  (including your CCM Administration credential). This password must be reset from the CLI and changed to something different before you can log into the system again. Note that this does not change your Prime License Manager (PLM) admin password if co-resident with your CUCM.

Security is important and continually making our security better is the only way to prevent incidents. With that said, proper planning will stop security changes from turning into outages.

Cisco’s TLS 1.2 Compatibility Matrix: https://www.cisco.com/c/en/us/td/docs/voice_ip_comm/uc_system/unified/communications/system/Compatibility/TLS/TLS1-2-Compatibility-Matrix.html

Advertisements

CUCM: The Logout Profile

Good Morning Readers!

Today’s post is quick and dirty. If you work in an environment where Extension Mobility is widely used, the content of this post is probably common knowledge and you can stop reading here if you like.

For those curious about Extension Mobility, there are a ton of resources on the Internet that can show you how to use and configure it as well as how it impacts your CUCM licensing. If you don’t know what CUCM stands for, this probably isn’t the post for you.

I was recently dealing some odd behavior from one of the 7841 IP Phones in my lab. Despite the fact that I had logged out my Extension Mobility (EM) user, their details (their User Device Profile) were still showing up on the physical phone. I reset the phone, checked the cluster for DB errors, and scratched my head a lot. I then proceeded to smack my head when I remembered the logout profile configuration in CUCM.

If you know anything about Unified Communications Manager Express (UCME), you know that the logout profile is very common on this platform. EM cannot exist on UCME without it. In CUCM we get spoiled, we can use existing device configuration details and do not have to specify a logout profile, there is however still a place for one.

In a perfect world a logged out EM configuration (in CUCM) should look like this:

This denotes that when a user logs out of the phone, the phone will use whatever configuration data exists in CUCM.

If however, a logout profile gets set (as shown below), the phone will display that profile upon an EM logout.

This is exceptionally confusing when the profile that was just logged out happens to also be the logout profile!

Obviously this was a misconfiguration on my part. Instead of just checking the box to enable Extension Mobility, I also specified a user profile. This lead to several minutes (a good half hour) of me troubleshooting a problem that inevitably had no problem at all.

Thanks for reading and hopefully you got some enjoyment laughing at my screw-up.

-J

 

CCIE Collaboration Lab: My Return Trip

If you follow this blog regularly (it says a lot about you, but that is for a different post) you know that last year, 2016, I took my first crack at the CCIE Collaboration practical lab exam. I took it in RTP (North Carolina) and it royally kicked my ass.

It is now nearly 12 months later and I am preparing for my second attempt. This year I have built my own lab instead of renting rack space from INE (who I believe I still owe money to from last year, but so far they haven’t come to collect). I am taking the exam on April 25 in RTP once again and while I am going to give it my very best shot, I am taking it because I honestly have no desire to attempt the written again and if I wait more than 12 months to attempt the lab my current written will be invalidated.

I’ll probably have more posts as the weeks draw closer, but the most important thing to note so far is that it is getting cheaper to purchase your own lab hardware. With Cisco coming out with the 44xx and 43xx ISR G3 routers the 29xx ISR G2 routers, which are still the hardware of choice for the current lab iteration, have become cheaper in the secondary market. I’m not saying that building a lab is cheap, by any means, but at least more folks now have that option. In my case,  my lab contains the following…

  1. 3825 – PSTN/BB Router. This is running CME for PSTN emulation. I am running PRIs to all 3 of the site routers; T1 PRIs the US sites and an E1 to the international site, I am using fractional PRIs to save on DSP resources.
  2. Dell Server 72GB Ram, Dual Xeon, SSD USB Drives (new addition) – This started life as a very weird CS-24TY but has now been revamped and runs all of my lab VMs easily.
  3. 2921 for HQ (2) PVDM-3 16 DSPs (this is actually enough for homogeneous video conferencing).
  4. 2821 for Site-B PVDM-2 64 DSP (good for voice only) (Generally Site-B is H.323 and possibly CUBE, a 2821 will run 15.1.x code which is not perfect but is close enough for that location). If I am asked switch/conference video at Site-B, I am S.O.L.
  5. 2811 for Site-C (CME)… I actually just ordered a 2911 and ISM-SRE-300 module to replace the 2811 as there are some serious differences between 15.1.x CME and 15.2.x and later CME i.e. CME 9/10.5/etc. I have a CUE module in the 2811 but I made the decision to spend money and get something closer to actual.
  6. 9971 phones as required (cheap enough used) and 7962 phones instead of 7965s as the differences between the two SCCP options are not that great and it saves some money.

In the back-end of the lab I also have a 3750 switch that I am using as a layer 3 WAN cloud instead of Frame Relay (which only matters in the QoS sections) and a 2960G PoE switch which I am using for phone power. I know the syntax is different but I cannot yet justify spending money or effort to aquire PoE  EHWICs.

I also have another Dell Server which hosts my “production” 11.5 CRS infrastructure which I can use for BB SIP calls as needed.

My setup is not perfect but perfect honestly costs to much. The fact that I can come down to my basement and practice whenever I want without having to reserve pricey rack space makes my setup perfect for me.

How about you? Do you have a lab? What are you running? Any suggestions? Comments? Questions? Leave them below!

-Justin

 

 

Cisco’s Virtual CUBE & Modern IOS Toll Fraud Security

Good Morning Web World!

I found myself with a little bit of time this morning and I thought I’d share a bit of my latest tinkering.

Those of you that have followed this blog for a while may remember my first post where I talked about pointing a CUBE through an ASA out to my ITSP, Flowroute. That post is located here for your reading pleasure.

While the software/hardware has changed with my setup the idea is basically the same. I still have a CUCM system (now 11.x) running with a phone (I’ve felt like being different lately so currently I’m using a retro 7985G as my endpoint (G in this case does not mean Gigabit)). I also have a firewall, a Cisco 5506-X (it was time for an upgrade from the 5505) and I do still have a CUBE. My previous CUBE was a 3825 and it worked wonderfully but the 3825 has long since outlived its relevance in today’s enterprise environments. In my stack of possibilities I also have a 2921 and while it is still a very powerful and valid router, it just seems too easy.

Simple and straight forward is great but only until you’ve done simple and straight forward, then it becomes time to mix it up.

To that end, my replacement CUBE is virtual. Yes, I said virtual. If you follow Cisco and their products, you may already know about the Cloud Services Router, the CSR1000V. The CSR1000V is a virtual router that runs on a VMWare ESXi host. It runs IOS XE though there is some Linux/Unix on the backend that makes it tick.

ESXI Virtual CUBE

Virtual CUBE Show Commands

 

 

When I first heard that it was possible to turn a CSR1000V into a CUBE, I was skeptical. As I have worked through the configurations and witnessed it work, I must say I am impressed.  The configuration is the same as with any other IOS XE router with exception of the interface naming conventions. There are three (3) Gigabit Ethernet interfaces and they are named Gigabit Ethernet 1, Gigabit Ethernet 2 and Gigabit Ethernet 3.  The configuration in ESXi is shown below.

CSR1000V Interfaces Config

In my case, I created three (3) separate interfaces on my vSwitch and pointed them at three (3) separate VLANs in my infrastructure. You could bond/Port Channel these interfaces if you wanted too but you will still be limited by the throughput of your host’s uplink(s).

A few important things to note…

  • When you deploy the OVA file in ESXi you are given a choice of multiple router “sizes” i.e. memory and processor. I am using the smallest size which is one vCPU and 4GB of vRAM. Some of the larger installations require an additional license.
  • Keep in mind that I am testing this configuration in a controlled lab environment. I am not sure of scalability and if you read Cisco’s CUBE configuration guide, located here, you’ll see that the virtual CUBE does have several restrictions that may make it impractical for some organizations.
  • The CSR1000V is a fully license-dependent platform. Once the demo runs out the router runs, but only with throttled performance. If you intend to deploy this solution, you will have to purchase licensing.
  • In my previous post, my 3825 CUBE was running 12.4 IOS which did not include Toll Fraud Prevention, starting in 15.x and moving through IOS XE you need to be mindful of TFP and what it means when you are trying to make/receive calls on a Cisco H.323 or SIP Gateway.

Let’s talk for a moment on Toll Fraud Prevention. As stated above, all modern IOS and IOS XE versions include Toll Fraud Prevention mechanisms and if you use a router, physical or virtual, for voice you need to be aware of them.

Voice Service Config

 

 

If you look at the configuration snippet above, you’ll see that in the voice service configuration there is a trusted IP Address list. That list contains the IPs of my ITSP and my CUCM. If I remove those IPs from the list, calls fail. If you are linking your CUBE or gateway to multiple CUCMs or other systems you’ll want to have those IPs in that list as well. What this list does is let the good/known IPs in to complete transactions/calls on the gateway/CUBE and keeps the unknown/bad IPs out to prevent them from placing calls on the system. IP addresses that are part of dial peers will be added to the list automatically but will not show up in the configuration.  If you’ve ever set up an IP PBX on internet, you know that Toll Fraud is a real and serious threat. If you don’t want or need this feature you can turn it off by entering “no ip address trusted authenticate” in your voice service configuration. This is not a recommended configuration but in your environment the TFP mechanisms may do more harm than good.

Thanks for reading, I hope this has been informative.

-Justin