dijetlo: I've told you in the past how easy KVM is to deploy and use. I've also told you the major enterprise users employ it around the world.

I think everybody in the business understands that there is some pullback on virtualization for services that are more appropriate for cloud and/or other more appropriate platforms. Not everything *should* be virtualized. So it's not a shock that some things will experience "un-virtualization".

Also, for the record, VMware is dead. Closed-source and a dead-end. I even work out with a guy who has worked there for a few years and he sees the writing on the wall. I used to prefer ESXi, but I've grown to appreciate KVM and RHEV over time.

VMware (and sadly HyperV) is king in the corporate world. It's far from dying. Companies want something with support and something that is turn key. Turn key being that you can install it and it just works, and you don't have to spend hours reading documentation (and trying to find it in first place) just to do something as simple as vlans or bonded networks.

Now personally I always found the whole support excuse to be BS, because when you call support you just get some guy in India that's googling for you anyway. Dealing with the complexities and expense of software licensing is just not worth it at times. But that's what the big execs want and they're the ones that make decisions, not IT.

In my case given there is a free version of vmware it was worth it. I would still be trying to figure out KVM.

One of these days I might look at KVM again to see if they made stuff easier and better documented so I can write a proper front end for it.

And based, in part, on that advice, I took the plunge and deployed QEMU-KVM.
By the way, two thumbs up, QEMU-KVM is very, very nice to work with, only took moments to initiate a virtual environment.
Now, getting an image I'm happy with has turned into a longer process but their are so many options available on the command line, it's gonna take time to work through them all.
A component of the cloud is virtualization, however I agree that inevitably when a new technology gets rolled out, especially if it has the ability to revolutionize a significant portion of development and deployment cycle, people go nuts with it. I think to some extent we're seeing that with virualization (or more accurately, we're seeing some of the old heads in these departments come to the realization "hey, we went a little to far wit this".
I wouldn't suggest virtualization wasn't going to be a big part of the future, I was only pointing out that virtualizing the little processor that deploys your cars airbag can have a significant downside.
Hmmm... I'm kinda of a fan of Oracles in much the same way I was a fan of Jaws. I'll believe VM is dead when Roy Scheider puts a hot round in the air tank jammed in it's mouth. Until that happens.... we should probably build a bigger boat.

You know Red, I really think that would be an appropriate little task for me and my fellow Qemu-KVM enthusiasts to help you with. You're exactly right, I found the fact that what seemed to be the intuitive "next step" often took you to another web site where you had to search for the document you wanted. Maybe a firefox "xml" file that could be imported into bookmarks as "Qemu-KVM Documentation" might not be out of line.
Anybody else up for it?

KVM is fully supported by Red Hat.

RHEV is enterprise-level and full supported by Red Hat.

VMware is the current king, yes. But not by much, and they are continuing to slip in marketshare all the time.

As for frontends for KVM...virt-manager is the standard front-end, and quite proper. Doesn't need any alternative, IMO.

I'm very happy to have made somewhat of a convert out of you.

Agreed.

I think you'll see VMware and others reach a parity where there is one or two kings (VMware, RHEV/Red Hat, Xen? (ugh)), and a few upstarts who can't quite cut the mustard, or don't have the enterprise-level features built into them for one reason or another (VirtualBox, KVM, ProxMox, etc.). We all know HyperV is toast at this point.

So when I say "VMware is dead", what I mean is that it is no longer "the king", no longer innovating, no longer saving money or providing the benefits that it once provided to their customers, and is instead just a run of the mill virt platform. That makes it vulnerable to losing additional marketshare to the other biggies in the market.

I would not call having to shut down a guest just to change the CD or network settings, proper. Virt-manager is a good start but it lacks a LOT of features. A proper VM solution/front end should enable you to do mostly the same things you can do to a physical machine such as hot plug network cable, USB, disks etc... At the very least, you should be able to change the CD (ISO) or network (change vlans etc) on the fly. Virt-manager does not allow this. Any change you make you get a message that the change will only take effect upon guest shut down.

Supposedly through virsh you can hot swap stuff though, but could not find proper documentation explaining the big picture, only little bits and pieces here and there and you have to figure out everything yourself.

If you are having problems with CD ISO mounting on KVM, then I think you're doing something wrong...

That's got nothing to do with virt-manager. Just because you scared me, I decided to test it to be sure and sho'nuff, I can mount, umount, change ISOs, remount and all works as expected. Same thing goes for the network hardware. You can add

Take a look at the type of hardware you are using, though. Almost any virtio-driven hardware can be hot-plugged, but IDE hardware, SCSI hardware, and some network hardware will require restarts. In my testing, I was unable to find a supported NIC type that required a restart for hotplug-adding.

Anywho, while I occasionally have to reboot when making changes, it's not very often. Not often enough to irritate me, at least.

Ah, if you don't want to read documentation, then go with Redhat.
They offer "turn key" solutions and have payed support.

No need.
If you want to make yet another libvirt front-end, you should start here.


Seriously, dude, dyasny is right: stop blaming technology for your lack of understanding of it.
KVM can be as easy pie or as difficult as rocket science, depending on what you want to do with it.

I have 4 windows server 2012 VMs and 1 windows server 2008, as well as 3 linux VMs, running in 2 hosts using KVM/libvirt.
Hosts are Slackware64 14.1 and are setup so that I have fail-over for all VMs.
I mostly use virt-manager for admin, but have some custom made scripts that help me automate things.
Speed for all VMs (windows based or not) is near native.
Disk I/O is the bottleneck, not CPU/RAM.

I've done nothing special to KVM to get it to behave properly. I installed it, installed the libvirt-manager packages and I'm GTG. This is on F20, so perhaps there are some OS and version differences, but otherwise, KVM is turnkey ready, IMO.

Oh it lets you change it, but the changes only take effect when you shut down the guest. There is even a message that says so. When I was installing windows 7 I had to put two cdrom drives, one for the windows 7 cd and one for the virtio driver cd, because I could not swap them in the middle of the install. So what happens if I need to install a distro that requires like 7 discs, I have to put 7 cdrom drives? lol. That is mickey mouse imo. I also sometimes need to swap vlans or even "unplug" a network cable on the fly. Can't do that either.

When I installed ESXi I was up and running in about 30 minutes and already had two test VMs going and was able to do all that stuff without any issues. I did not even have to dig for any documentation because it is designed to be simple to use. if I need to do something very advanced then perhaps I will need to google it. But I should not have to google simple things like how to setup vlans.

This was not the case in my environment. Sorry. I had not issue switching and *using* the switched iso files. Perhaps it's the way it handles Windows?

Of course you should have to RTFM for advanced ("simple"?) things like vlans...

[QUOTE=Red SquirrelOh it lets you change it, but the changes only take effect when you shut down the guest. There is even a message that says so. When I was installing windows 7 I had to put two cdrom drives, one for the windows 7 cd and one for the virtio driver cd, because I could not swap them in the middle of the install. So what happens if I need to install a distro that requires like 7 discs, I have to put 7 cdrom drives? lol. That is mickey mouse imo. I also sometimes need to swap vlans or even "unplug" a network cable on the fly. Can't do that either.
[/QUOTE]

you can change it on the fly, might be you're using a version of virt-manager that is lacking, but that's virt-manager, not the hypervisor. Try to install oVirt and use spice -you'll see how much better virtualization can be, than anything you'd get out of ESX

1 members found this post helpful.

After thinking about this, I wanted to mention what I believe the problem to be.

If your implementation of KVM requires that you load VirtIO drivers, then it stands to reason that your hardware will not behave until after they are loaded.

I once again fired up a test VM-this time, Win7-and mounted, unmounted, used, accessed hardware and ISOs with no problem. I want to note that I have *never* had to load VirtIO drivers on any of my KVM instances to get the hardware to behave. I'm not saying there aren't situations where they are required, but my experience tells me that unless you have a very old version of KVM and virt-manager, that this is not a problem.

Either you misunderstand what you're doing (which I honestly doubt), or there's something wrong with your setup.

"my setup" was "apt-get install kvm" on a completely fresh install. (or kvm-common I forget exactly what I had to type) I don't see what I could possibly have done wrong, it was as vanilla as it can get. Everything "worked" out of the box so to speak, it just did not work as well as vmware does, and had lot of quirks that had very complex fixes like editing xml files and crap... never got that far, as it was not my idea of how I want to use a hypervisor. Anyways, I'm very happy with ESXi so I will most likely stick with that anyway. My only fear is that it ceases to be free or they start adding more limitations at some point. I'll cross that bridge when or if it even happens.