Upgrading 9.1.1 to 11.5 with Refresh Upgrade

Mike Elliott · ‎08-30-2018

So I have a UCS server (220) that has a standalone CUCM, IMPS and UCXN server on it. We are in the process of testing upgrading the plant. Since it is stand alone, with no subscriber nodes for any of the services, what I did was cloned the VM's to a new ESX host, reconfigured the VM Network to be on an isolated VLAN and then booted each machine. I installed the new RSA Keys on each machine then made sure I had enough disk space to run the upgrade.

I ran the upgrade on each cloned VM and everything seems good. The problem is, I cannot test this against our gateways (isolated vlan), nor can I test registering any phones or Jabber clients.

My plan is to test a cutover next Wednesday morning. My approach is to go to the C220 and disconnect the network on each of the 9.1 VMs then connect the network on the CUCM, UCXN and IMPS 11.5 servers.

Does this approach seem like it would work? What gotchas should I look for? Is there anything I should check on the 11.5 servers before attempting to go live?

The concerns I have are these:

1) Our CUCM/UCXN gets it's authentication for our AD. I do have control over the AD so I am not so worried about it not being able to connect, but is there anything new in the Jabber config xml files that I should be aware of?

2) Our inbound call flow works like this

Call comes in from PSTN via a 2911 and gets forwarded to CUCM

CUCM forwards call to UCXN to go to auto attendant

UCXN checks time of day and holiday schedule

If the call is during working hours and no holiday it forwards it to a hunt on CUCM and goes to admin desk phones

If the call is out of hours, it sends it to a voicemail tree

There are only two inbound numbers, our main number and a support number. We have no DID, all calls are handled by UCXN or admin phones and transferred accordingly.

We also have a trunk setup to a CME in one of our branch offices.

All of our phones are 6941, 7942 or 7962 with cap units

We have an Informacast paging server but it will not be upgraded in this round.

Thanks for any advice you can give.

M

Jonathan Schulenberg · ‎09-02-2018

What you’re describing is also referred to as an “island mode upgrade”. The way you _should_ take that approach is to emulate at least DNS and NTP in that island. A Cisco router can act as both; just configure loopbacks on it with the IP addresses of the DNS and NTP servers and make it the default gateway. It’s in the “island” along with the rest of the stuff but NTP and DNS resolution will appear to work. (The firewalld tables get populated at boot based off DNS resolution.)

Jabber config has changed substantially but that is mostly client-side when you update the clients. Read the release notes for the required transition from BDI/EDI to CDI directory integration in Jabber 11.8+ before updating the client.

Make sure you verify Informacast compatibility and whether or not it’s using the TSP. If it is the TSP will have to be updated for sure.

Mike Elliott · ‎09-05-2018

So it does look like the plan worked. I did not manage to test AD/NTP before hand so it was a total unknown. We're only a few hours into the new system to know if there are any real bugs.

We're running on the 11.5 plant now. The only issue I had was that the servers did not immediately have network access. I had to change the VM Network they were on, then change it back to get it going.