Migration from VMware EsX to Redhat KVM


Can anyone help me with pointers on the following:
- Migration approach from VMware ESX/ESXi to Redhat KVM  for Prod and Dev environments
- Risks that need to be addessed and mitigation plans
- Business benefits of migrating to redhat Kvm

I have gone through some Redhat docs , but searching out for tested approach with respect to both Prod and Dev environments along with the business benefits that can be derived out of this migration.

If anyone has any docs/articles/ppt , these  would be of great help. need this information urgently.


Who is Participating?
Migration approach from VMware ESX/ESXi to Redhat KVM  for Prod and Dev environments

First, what Guest OSes do you have that you want to migrate?

Next, what is your scope? Are we talking about a handful of VMs with no need for an equivalent to VMotion? Or do you want to implement an entire KVM Farm with shared storage, HA, etc?

Risks that need to be addessed and mitigation plans

Risks? VMware ESX is essentially RHEL, stripped down, with a custom-developed management software layered on top. So, in that sense, moving to KVM is really just swapping out the virtualization management layer.

Also, any examination of risks needs to touch on alternatives - the other big virtual player in the Linux world (beyond VMware and KVM) is Xen. Personally, I prefer KVM to Xen because KVM is just another kernel module - insmod and you're in business (OK, it's more complex than that). While Xen has some neat things, it is, unfortunately, a special kernel build. ISVs will latch onto that in order to shift blame. If you run into a problem, they'll demand you replicate it on a "standard" kernel build before they'll investigate it. People can argue to technical pros-and-cons of Xen vs. KVM, but for my money, the architecture of KVM will give fewer headaches in the long run.

Now, to mitigations. First, I'd do a proof-of-concept. Stand up a KVM Host. Now build a simple typical Guest - what do you need to do to execute your standard system build process? Does it work reliably? Are you going to have to create an entirely new system build process?

Second, P-to-V (or V-to-V) a not-mission-critical system. Maybe a redundant caching DNS (not your primary DNS server!), or something that's similarly small and simple - but used sufficiently so that if there's a problem you'll actually find out about it. Run it awhile. Do your monitoring and management tools get along with the new underlying platform (obviously, if you do a V-to-V, VMware-Tools won't like KVM)?

When you've done those things, you've taken several large steps towards determining and mitigating the risks in your environment.

Business benefits of migrating to redhat Kvm

The largest business benefit is not having to pay the VMware Tax for your VM environment. If your Guest systems are RHEL, then if I'm not mistaken, you can host a (theoretically) infinite number of RHEL Guests without consuming a support entitlement for the Guests (the Host or Hosts still each require a support entitlement).

Another benefit is breaking free of Winblows. VMware pretty much forces you to use Winblows to manage the VMware environment - the VIC interface is tied to Internet Exploder, and even the web interface to the VIC doesn't officially support anything else (you can get most things done - remote console to the virtual host won't work, because it's tied to I$E - but if you run into problems, forget about getting any support from VMware once they find out you're using Firefox or a Linux-based client).

Other business benefits are likely specific to your environment.
uninstall vm tools, convert exported image and attach to kvm...
abbis_inAuthor Commented:

Thanks for the details . It's certainly helpful.

To answer your querie , the guests OS are predominantly Windows server 2003  and some  2008 that needs to needs to be migrated, also there will be a requirement to implement an entire KVM Farm with shared storage, HA, etc as this migration is targetted for Prod environment as well. So was looking for a project/implementation plan of exactly how this migration need to be approached assuming that the existing storage would be utilised.

Well, I suppose that means you won't be escaping Winblows anytime soon. :-)

You're going to need some clustering add-on to get HA. That is, there's nothing inherent in KVM that will provide HA, and while you can engineer the equivalent of VMotion, I'm not aware of any vendor offering a KVM solution that does that sort of thing out-of-the-box. I know RHEL has a Clustering/HA offering,  but I've not played with it.

One issue you'll need to deal with early on is the filesystem for the shared storage. Can't just slap down an ext3 (or ext4) filesystem and share it. You're going to need something designed for sharing - perhaps OCFS2 (which I've used - I don't like the kernel-specific builds, and Oracle simply doesn't get Linux no matter what their marketing drones say, but it does work and is designed for filesystems to be shared across multiple nodes) or GFS (not tried it).
abbis_inAuthor Commented:
It was very helpful.

Question has a verified solution.

Are you are experiencing a similar issue? Get a personalized answer when you ask a related question.

Have a better answer? Share it in a comment.

All Courses

From novice to tech pro — start learning today.