Page 6 of 8

Re: Virtualization inside a virtualizer, discussions

Posted: 22. Aug 2013, 12:07
by Ramshankar
Technologov wrote:The good news, is that the new CPU, Intel Core i7 4770 "Haswell" supports nested hardware virtualization (VMCS shadowing).
So it would be good to enable this new feature in VirtualBox.
As I mentioned in the other ticket, yes we are very well aware of what Haswell supports, but at the moment it looks like if and when we do support Nested virtualization it will probably have to work on CPUs older than Haswell as well.

Re: Virtualization inside a virtualizer, discussions

Posted: 2. Oct 2013, 11:09
by more-solutions
Slightly different use case here; reading this thread I think it'll work but comments welcome..

I am doing some work on a Linux distro/appliance that will have a plugin to control VirtualBox. I need to play with the plugin code.

I already have VirtualBox on my Ubuntu 64-bit host. I want to create a VM in which I will install that distro, and within it I will then need to install and manage VirtualBox.

All I need to be able to do within the virtual environment is create VMs and be able to start them etc, I don't really care whether they get past POST and I have no particular desire to install anything on them, I just need to test the plugin's ability to manage VBox. (Any "real" user of the distro is going to install it on real hardware.)

It sounds like this will be fine?

Re: Virtualization inside a virtualizer, discussions

Posted: 8. Oct 2013, 20:58
by lamont
Found this thread due to a simple use case I'm working on right now that sounds similar/identical to the previous poster.

I'm writing a Chef cookbook that will put vagrant+virtualbox+docker onto an Ubuntu system image. The goal is to get cookbooks suitable for setting up servers that have a base config which I can run CI/CD infrastructure on. These servers will eventually run something like jenkins which will be spinning up virts and testing (other) cookbooks and seeing if they pass or not.

In order to quickly write and test and debug these cookbooks I want to do development work on my mac laptop, but docker only runs on ubuntu so any idea of running the cookbooks I'm testing on my mac image is a non-starter and isn't what I want for other reasons as well. So I want to be able to spin up an ubuntu image, run chef cookbooks that install vagrant+virtualbox+docker, and then validate that those are running correctly.

This is for development purposes, but eventually these servers should be doing CI/CD on their own cookbooks. They should be spinning up new virts, checking development branches of their own cookbook, and when those tests pass the cookbooks should get promoted and then applied to the servers themselves.

This is becoming a perfectly reasonable thing to do, and there are companies building infrastructure like this, and the ability to do virtualization-on-virtualization like this is going to be useful to get complete testing of a companies entire infrastructure. It isn't going to be necessary to get it running fast, and nobody would use this for persistent production instances, but for dev and Q/A it makes sense to do virt-in-virt testing.

Since this is going to be a narrow set of servers that need to use it, having it only work for Haswell would be acceptable. If a company does this, they only need to have one team with Haswell laptops and have a few build servers be on Haswell for virt-in-virt testing. There's no reason (at least based on this use case) to require the entire infrastructure be on Haswell, or to require virt-in-virt across the entire Enterprise.

Re: Virtualization inside a virtualizer, discussions

Posted: 11. May 2014, 23:58
by blong
Anyone running VirtualBox inside VMWare Fusion?

Re: Virtualization inside a virtualizer, discussions

Posted: 12. May 2014, 00:14
by loukingjr
blong wrote:Anyone running VirtualBox inside VMWare Fusion?
I think a better question is, why would anyone want to? You have the overhead of running Fusion, then the overhead of running a guest VM in Fusion, then the overhead of running VirtualBox on the guest VM, then the overhead of running a guest VM in VirtualBox. No thanks.

Re: Virtualization inside a virtualizer, discussions

Posted: 12. May 2014, 00:44
by blong
A fair point @loukingjr :) . The reason is to prepare a host that will then be moved to vSphere. Although, now I'm thinking I may be better off starting this experiment in vSphere rather than trying to export the configured host from VMware Fusion to vSphere.

Re: Virtualization inside a virtualizer, discussions

Posted: 12. May 2014, 00:48
by loukingjr
blong wrote:A fair point @loukingjr :) . The reason is to prepare a host that will then be moved to vSphere. Although, now I'm thinking I may be better off starting this experiment in vSphere rather than trying to export the configured host from VMware Fusion to vSphere.
I try :wink:

Re: Virtualization inside a virtualizer, discussions

Posted: 12. May 2014, 00:51
by blong
@loukingjr Would you say configuring the host in VMWare Fusion is a bad strategy? Think I should start in vSphere first?

Re: Virtualization inside a virtualizer, discussions

Posted: 12. May 2014, 00:53
by loukingjr
blong wrote:@loukingjr Would you say configuring the host in VMWare Fusion is a bad strategy? Think I should start in vSphere first?
I'm really not familiar with vSphere. I also don't use Fusion. You should ask on the VMWare forums.

Re: Virtualization inside a virtualizer, discussions

Posted: 19. Aug 2014, 15:03
by AnrDaemon
Anyone had a success using LXC under Linux guest?
Moving forward to a 64-bit host OS and better utilization of RAM, I'd like to start some heavier experiments from my list of "would be nice to do" things.
I know it's not quite the virtualization, but close enough to ask before try. :roll:

Re: Virtualization inside a virtualizer, discussions

Posted: 24. Nov 2014, 21:01
by walkamongus
I'm running a Virtualbox VM guest inside a Virtualbox VM host (via Vagrant) and the VM-inside-the-VM boots up fine. The issue I'm running into is that I am unable to access the VM guest on the host-only network from the virtualized host. The vboxnet0 interface is up in the virtualized host and it has the default gateway IP address that I can correctly ping, however I am unable to access the VM guest that is on that network.

I tried this setup by virtualizing the host on both VMware and KVM and had the exact same issues, which is why I think the issue is not running Virtualbox inside Virtualbox.

Has anyone encountered this issue before?

Re: Virtualization inside a virtualizer, discussions

Posted: 25. Nov 2014, 11:49
by mpack
I'm afraid we don't support Vagrant here (the VirtualBox forums).

Re: Virtualization inside a virtualizer, discussions

Posted: 13. Feb 2015, 14:30
by Legorol
I would also love to see VirtualBox support nested hardware virtualization. I don't mean the new features in Haswell CPUs, but virtualized VT-x that would work on older CPUs as well. The aim is to be able to run hypervisors inside a virtual machine.

The use case that I have is setting up test/lab computers for virtualization itself (for example, Hyper-V). Currently such test cases can't be set up inside VirtualBox. VMware Workstation supports virtualized VT-x, so I tend to set up test cases there. I much prefer VirtualBox to VMware Workstation and run all my VMs in VirtualBox, except when I need nested virtualization.

Re: Virtualization inside a virtualizer, discussions

Posted: 13. Feb 2015, 14:38
by loukingjr
@ Legorol, Is VMWare Workstation free now? (rhetorical).

Re: Virtualization inside a virtualizer, discussions

Posted: 13. Feb 2015, 16:18
by mpack
Legorol wrote:virtualized VT-x
Do I gather that you mean a fully synthetic VT-x that works on CPUs that don't have it? I can't speak for the devs but... that ain't gonna happen. The whole point of VT-x is to do things that you can't do efficiently in software alone. It wouldn't need to exist otherwise. That and CPU's without that feature form a rapidly shrinking market of penniless users... hence ISTM targetting them is not a sensible use of resources.

OTOH, if you want to do that work yourself and then donate the code then that would certainly turn my calculation upside down.