RHEV 3.1 – Power Management for Cisco UCS Blades

Hi folks,

Here’s my next info blurb about RHEV 3.1 – this time it has to do with “Power Management” configuration for a Cisco UCS Blade. It’s another case of initial frustration until you finally figure it out and it finally makes sense. It’s also a case of “there really isn’t much help out there”…

Ok, here’s the full situation: you have RHEV-M and NetApp storage deployed. Now you need to add in some hypervisors. You’ve been given access to the latest & greatest Cisco UCS Blades (B200 M3). You deploy RHEV-H on the blade, utilizing the NetApp for SAN boot, and even “approved” the UCS blade in RHEV-M. Everything else (Data Center, Cluster, etc) is configured; all you have left is the Power Management configuration.

What’s so important about the Power Management anyway, you ask? Simply put, it’s what puts you back in business when your hypervisor is impaired enough to degrade your virtual machines, but not enough to crash. Essentially, you’re in a state of limbo that requires a hard reboot in order to release resources like virtual machines and storage domains. The best example is that the physical server itself is in a hung state. If it’s the “SPM” in addition to hosting numerous VM’s, you’re in a pickle.

The Power Management is what allows you to force the blade to reboot directly from RHEV-M, thereby releasing the storage domain to another hypervisor in the cluster. If the VMs that were running on the hypervisor are configured for High Availability, then they will restart automatically. Regardless, you’re back in business without having to locate the specific server in your data center.

Ok, so let’s get to brass tacks here and configure this.. The configuration itself is really easy. There are 2 parts: configure a user in the UCS-M that has power access, and configure the Power Management for the RHEV-H node.

Assuming you’re already logged into UCS-M, click on the “Admin” tab, then use the drop down next to “Filter” to select “User Management”. Right click on “Locally Authenticated Users” and select “Create User”. In this case I used the user name “test”.. I’m sure you can come up with something better or more creative. Be sure to enter a password, and under “Roles”, be sure to select “admin”. Click on “Save Changes”. Lastly, note the “Service Profile” that is in use for the UCS blade in question.. In my case, it’s called “rhev”.

UCS_user

Next, we’ll configure and test the Power Management in RHEV – super easy. Right click on the hypervisor, click “edit”, and then select “Power Management”. Check the box that says “Enable Power Management”. The fields should be filled in as follows:

  • Address – IP or FQDN of the UCS-M
  • User Name – the user name that we configured above just for power management
  • Type – select “cisco_ucs” from the drop down
  • Slot – this is misleading in that we don’t want a “slot number” for the blade, we enter the “service profile” name
  • Options – leave blank
  • Secure – you must check this box or the test (and actual power activities) will fail. Miserably.

When you’ve filled it all in, click “Test”. You should get a success message that looks remarkably like mine.

rhev_power

That’s it, you’re done!! But what is Captain KVM doing with UCS Blades, and Nexus Switches, and NetApp storage, you ask.. GREAT question. I can’t give you the answer. Yet. 😉

Hope this helps,

Captain KVM

4 thoughts on “RHEV 3.1 – Power Management for Cisco UCS Blades”

  1. What documentation did you use or come up with for the “Everything else (Data Center, Cluster, etc) is configured”

    Finding this documentation to be a little sparse out there.

    1. Hi datadudewayne,

      If you’re talking about RHEV info specifically, go to http://access.redhat.com, and look at the RHEV info under documentation. They might still be running a free 60 day eval. Or you could download the upstream oVirt version free and clear. If you’re talking about Cisco UCS, the info is a little sparse. Do a google search on “flexpod, rhel” and you should find a Cisco Validated Design document that talks about best practices.

      hope this helps,

      Captain KVM

  2. Thank for sharing but it’s not working in ovirt 4.
    i did as you instructed but when i hit test it just fail to connect to the ucs-m.
    any thoughts Captain ?

    1. Hi Mahdi,

      There could be a firewall blocking access. Can you reach ports 80 and 443 from the ovirt engine? Another possibility is that UCS may have been updated since I created that post (3.5 years ago) and there may be other steps necessary on the UCS side to allow access. Your troubleshooting should start with port access (80, 443) from the ovirt engine. Then login/password, then profile, etc. The actual power management/fencing mechanisms are straightforward.

      Captain KVM

Agree? Disagree? Something to add to the conversation?