Cisco Unified Communications VMs Support

[k00laid]

We run a good bit of the Cisco Unified Communications platform inside of VMware here and one of the things expressly forbidden by Cisco is using Image Based backup utilities with the systems, going so far as expressly naming Veeam. As long as they say it is unsupported we can't extend Veeam to backing them up, any chance you guys can work with Cisco and get your product supported?

Here's a link to the documentation:
http://docwiki.cisco.com/wiki/Unified_C ... 8.6.281.29

[tsightler]

So I think you should reach out to Cisco on this, there's nothing specifically that Veeam needs to do.

Cisco would first need to support snapshots. This is unlikely since, the primary reason they do not support snapshots is the time sensitive nature of their application and the impact that snapshots can have on the underlying system performance. This is more critical for voice/video situations than for most general purpose servers.

Also, Cisco would ideally need to improve their database replication to automatically recognize systems that were restored from snapshot and resync them. Right now this requires manual intervention.

Technically, you could use Veeam today, but you have to power off the VM. If you look under "Copy Virtual Machine" section it specifically states that copying a VM is supported as a backup option as long as you power off the VM (and they also note the caveat of manual intervention being required to fix database sync). So, if you power off the VM, and then back it up with Veeam, you should be OK, since basically all that Veeam does is make a copy. What they specifically do not allow is snapshots of a running VM due to the potential service impact (there's no service impact to snapshotting a VM that is powered off), so when they refer to Veeam specifically I'm sure that's what they are referring to as they support no snapshot based backups of running VMs because they do not support VMware snapshots at all.

[chrisdearden]

What tom said I deployed a bit of US on UCS a while ago and it was hugely prescriptive about what you could and couldn't do at the time. I'm sure they will relax them a little bit, though as I understand it, an offline backup of the VM is the only supported method for maintaining availability outside the HA features built into the UC applications themselves.

[bbricker]

Sorry to bring an old thread from the dead but wanted to add to this. Now that Cisco has relaxed their rules on running their UC vm's on non-Cisco servers, I am thinking of moving them off my 2 UCS servers that are running ESXi Free and put them on my larger Dell vmware cluster. This means I could also back them up with Veeam.

Is anyone out doing this with success? I don't care if it makes the UC apps unavailable temporarily while it snapshots since it will be happening late at night, I just want to make sure it doesn't do any "damage" so to speak. Even if I had to restore one of the UC vm's and something wasn't DB consistent, I would still have my regular/traditional Cisco backups that run out of each app nightly.

[Dima P.]

Hello,

That is correct - Cisco unveiled the possibility to set up the UC on the VM sometime ago - and a short look across the Web does not bring any mentioning of the serious issues that may arise during set up/migrations or evaluation of such set up.

For the detailed explanation of the setup best practices - Cisco or VMware community is the best place to find the answers. Anyhow if you have existing UC servers well-functioning - you may set in VMware a couple of test ones and see how they would behave. In case of success - complete the migration with eliminating the old UC servers.

Please do not hesitate to update this post with the results of your testing,
Cheers!

[bbricker]

Thanks for the reply but I am interested in other Veeam users who are doing this and their experience.

[k00laid]

Sad

Thanks for the reply but I am interested in other Veeam users who are doing this and their experience.

Sadly this looks like it still isn't supported. I was at Cisco Live last week and specifically asked one of the product engineers about it, it has something to do with the way the database stays open. His recommendation regarding using Veeam to help support a disaster recovery plan was to shut all the components down on a weekend and hit them with a Veeam job and then continue to run DRS. If you have a backup job with the latest version of whatever UC components you've got with current DRS backups he said expected recovery time for my environment (4 components x2 for HA) would be about 1.5 hours.

Here's the latest Docwiki showing it is still a no no.
http://docwiki.cisco.com/wiki/Unified_C ... ckup_Tools

[veremin]

As far as I know, the major problem of backing up/replicating VMs that run Cisco specific application is related to the sensitive nature of such applications. In other words, the traffic going between Cisco VMs is supposed to stay real-time and the usage of backup application leads to certain traffic drops. Therefore, it’s always recommended to backup these VMs in shutdown state.

However, assuming that there is a time interval when you’re ok with shutting down these VMs, you might as well back them up in power on state.

Thanks.

[tsightler]

Actually, the time sensitive nature isn't so much of a problem if they are clustered, as they are in most deployments. The real problem is that there is no supported method to restore these systems once you back them up. I believe this is because of the replication used between the nodes to keep them in sync. If I suddenly restore one of the nodes to a prior point in time the other node is not aware of this and the databases will begin to diverge. It's similar to the problem with DSN rollback on AD. Effectively, the databases assume that once you committed and acknowledged a certain transaction, you will not suddenly be transported back in time an no longer have that transaction.

I suspect that you would always be able to restore the Veeam image and then restore the DRS backup, since this process takes care of that issue. I've always thought that an option for simple 2 node deployments would be to backup only the publisher/primary node with Veeam. That way, after a disaster, you could simple restore the publisher node VM but you'd have no subscriber nodes. This would be enough to get you back working very quickly, and then you can always manually redeploy a new subscriber node later. Probably doesn't work for larger deployments when many subscriber nodes are required to support the client base, but in those cases it's unlikely you'd ever loose the entire environment anyway. Unfortunately I no longer admin such an environment so I've never been able to test any of this.

[jarein]

Hi all,

I see that many guys are interested in this subject. I've tryed to use veeam to backup a CUCM VM on UCS, with Cisco license on ESX 5.1, instead of using the esx embedded VM copy. The final issue, even with a VM shutdown, was that Veeam was providing an ESX license issue. As we were using the Cisco delivered license as usual, as Cisco integrator, we are supposing that Cisco was modifying the licensing content, and die prevent the use of expected services for the backup to work.

Our target in the lab was to check if with the supported method (VM shutdown and copy/export), Veeam could help to speed up the process.


It would be interested to know if Veeam could progress or talk with Cisco in order to have a common business to fullfill here. I'm suspecting Cisco to try to bring the UC customers to have the full UCS Datacenter design, with Nexus, Storage, and so on.

[dwerts1]

I see there is a thread a little over a year old on this topic. Wondering if anything has changed as to the backup procedure for these Cisco VMs or if anyone has any scripts etc that work

[veremin]

Nothing has changed in terms of specific integration with the said solution. So, the answers provided above should still be valid. Thanks.

[Manny15]

Would be nice if Veeam could restore the backup image, then also restore the DRS backup...

[bilalzaidi]

so this thread is too old now

[k00laid]

Yep, essentially the problem isn't Veeam, it's Cisco that prohibits any 3rd party from backing up their systems while live. FWIW the work around I use keeping a manual Veeam backup job with all my Voice VMs in it and after I complete any systems maintenance on them I shut them all down and take a backup which can be shipped anywhere. I then also have a little Linux VM I use to be the repository of DRS backups that can then be shipped to the same places. Our DR work flow for this is to restore the VMs from Veeam backup and then layer the latest DRS onto it. As all the VMs are there in backups this can even be done in a SureBackup lab for testing of either DR or to dry run Cisco patches on the system.

[ucengineer]

Hi, I have to move my CUCM publisher into the different dc and only veeam is the solution but if I migrate our publisher, what kind of changes I need to take care of? About the subscribers, do i need to add them again or will there be any problem after that?

[jhoughes]

In a previous role, we ran into issues on many older versions (v6-v11) with the migration of publishers having the potential of subscriber registrations became invalid when the publisher first came up. It was sporadic, and sometimes it was resolved with reboot or writing only the subscriber configuration/registering again. We even had a few instances of the publisher itself reporting corruptions or having config issues at initial boot which required manual correction at the console.

After many months of internal work (Cisco gold partner & #1 global provider of CCaaS for Cisco) and work with Cisco up to v10, we just got to the point of migrations becoming array-level replication jobs of the VMs themselves to re-import into vCenter inventory, and our failback plan was always new publisher deployments with recreating the configurations if necessary. We had the benefit of quite a few years of developing our tools to be able to properly mirror or full recreate the configuration, even when using some unsupported methods, and Cisco supporting us in that effort. Sometimes, standard TAC may not support you very much on migration and will tell you to redeploy, or will only assist in getting it back up in the original location/configuration.

I don't mean to sound overly ominous about it, but I've had many hours/days/weeks of trouble as migrations could have the potential for hiccups or full disaster, and I'm not certain how much Cisco would support issues that may be within your virtual environment.

[k00laid]

Hi, I have to move my CUCM publisher into the different dc and only veeam is the solution but if I migrate our publisher, what kind of changes I need to take care of? About the subscribers, do i need to add them again or will there be any problem after that?

One option if you want to use Veeam to do your migration would be to shut the VMs down and then hit them with Veeam. You can do that with any of the virtual servers, you just can't hit a running Cisco IPT server with Veeam and not violate the support terms. Everytime we upgrade CUCM I will shut them down immediately after and then hit them with a Veeam job just I'll already have a proper VM to restore DRS to if need be.