Future of virtualization

[SamTzu]

Hi, all.
About every year I take a look at virtualization options and this time I think I see a clear trend forming. KVM and LXC are becoming more and more popular partly because XEN and OpenVZ are lagging behind in the development. Reasons for this are fairly obvious. KVM and LXC install/configure easily and work both straight without any kernel games and come supported by both the major Linux groups (.rpm & .dep repositories).
This can't be said about XEN or OpenVZ. We all saw what happened with OpenVZ support on Proxmox last year. I believe this is just beginning signs of the end of OpenVZ development. Let's face the facts here. We are all lazy admins (more or less.) Everything that requires an extra effort comes that much harder for us. We would rather not put in the extra effort if an easier solution presents it self. Hey, that's the reason we love Proxmox. It's super simple to implement and use. I have been looking in to OpenQRM because it supports LXC. I would very much like to move to LXC from OpenVZ since LXC does not require modified kernel and it's memory management is way better (no more complicated beans on my coffee.) OpenQRM is super complicated though. It takes days just to make it work and weeks to to be able to learn how to use it. Not anything like Proxmox. Fact is OpenVZ is not safe to use in a meaningful way anymore. Too many linux components require CAPABILITY="Security Hole/MySQL injection vulnerability" to get it to work on OpenVZ these days.

All in all I think the signs are clear. We are going to see XEN and OpenVZ starting to fade away. What does this mean for us? Who knows. My money is on LXC. I for one do not intend to move our OpenVZ containers to KVM.

 

[SamTzu]

Flame away

 

[tom]

OpenVZ will be stable and supported in the long term 2.6.32 (RHEL6) branch. I do not see a huge run away from OpenVZ to LXC in the moment (still missing features).

 

[SamTzu]

I have heard that before. I'm not buying it. OpenVZ is not RedHat Linux but rather RHEL is dependant on OpenVZ project's support. If that support stops then we will see RHEL dropping it's support also. Missing features and almost non existent documentation on Grub2 did not prevent almost ALL the major distro's going with it anyway. I predict we will see the same happening here.

 

[dietmar]

Missing features and almost non existent documentation on Grub2 did not prevent almost ALL the major distro's going with it anyway.

Well, grub2 is a mystery too me to. Anyway, LXC is the future, but current production systems use OpenVZ.

 

[SamTzu]

Yep. It seems so. That's why I'm studying LXC. I have nothing against OpenVZ or XEN. They have their problems like most APP's do, but I'm a firm believer of the old KISS principle. "Keep It Simple Stupid."
So far Proxmox has done a very good job of applying that principle. I'm hoping that you guys keep up the good work and keep it simple.

 

[naturlight]

I'm using KVM for all my VPS < 400 with proxmox and it's rox !!
No crash, it's working very fine and support windows and linux.
Thanks to proxmox !

 

[FlappySocks]

I'm new to OpenVZ and LXC. Mostly been using Xen/KVM/Virtualbox for the for the past 4 years. I'm about to set up 3 new servers for virtualisation of 5 or more Linux instances each, and so in my research stumbled across Proxmox & openQRM.

So having never used OpenVZ/LXC & Proxmox/openQRM, i'm unbiased at the moment as to what to use.

My feeling at the moment, is as SamTzu said, Proxmox is KISS, and openQRM looks more complicated for little extra gain. However, openQRM supports LXC, and I feel a lot more comfortable with that in the long run (maybe incorrectly) as its supported in the upstream kernel. What exactly is LXC missing?

 

[SamTzu]

Well, most people only need the basic stuff which the LXC all ready has. What it needs is a web based control panel like Proxmox's for OpenVZ and KVM. Fact is that the most used buttons on Proxmox are start VM, stop VM. Suprise. The migrate button belongs to the "advanced" category. I have been playing with OpenQRM now for over a week and I'm about ready to give up. I only looked at it because I too feel that LXC is the future of Container based virtualization.

 

[FlappySocks]

Maybe if proxmox uses libvert, switching to LXC is trivial?

 

[dietmar]

Maybe if proxmox uses libvert, switching to LXC is trivial?

Switching to LXC is trivial anyways - I guess I can write up something usable in one week. But LXC is still missing some features compared to OpenVZ (live migration, security). My plan is to give it another try when we have 2.0 stable.

 

[FlappySocks]

Another thing that bothers me, is finding a hosting provider that will install proxmox on a dedicated server for me without charging a lot to do it, and then washing all hands of responsibly.

How cool would it be, if I could just rent any dedicated server, with a modern distro, install a .deb or wherever, and start using proxmox. Add ksplice, and you wont have to touch the kernel then

 

[tom]

http://forum.proxmox.com/threads/4638-Recommended-Proxmox-VE-Hosting

 

[ichilton]

Hi,

Interesting post but please could you clarify a few things?

We all saw what happened with OpenVZ support on Proxmox last year.

What happened?

memory management is way better

In what way?

Too many linux components require CAPABILITY="Security Hole/MySQL injection vulnerability" to get it to work on OpenVZ these days.

Please could you explain (or give url's) what you mean by this?

Thanks,

Ian

 

[FlappySocks]

I do not see a huge run away from OpenVZ to LXC in the moment.

This may be true, but I think there is a huge untapped market.

There are only two hosting providers offering Proxmox, where as just about every hosting provider can supply you with a dedicated server able to run a distro with KVM & LXC already built in.

As a solution provider, I buy up lots of VPS instances for my customers. For me there is no middle ground between dedicated servers and VPS's. It would be cheaper and more flexible if I could rent a bunch of dedicated servers, and manage my own VPS's, with the knowledge that if the hardware fails, the hosting provider will take a look at my "debian server" or my "CentOS server", or whatever. They have no need to know I run proxmox on it. And if I do need proxmox support, I would rather pay Proxmox Server Solutions who have an interest in the product.

 

[aub]

I would like to warm up the discussion on LXC in Proxmox.

Someone please correct me if I am wrong about this but I recently ran into unfortunate circumstances with an OpenVZ Ubuntu 10.04 setup.
I like OpenVZ because it's so easy to set up. But after running into severe issues with ip_tables/post-routing, I saw some posts that suggested Ubuntu doesn't support OpenVZ in version 10.04 and later. It appears Canonical is betting on LXC. I had to revert to full virtualization (KVM) to get ip_tables/post-routing to work since I didn't want to, if I do get this right, recompile the 10.04 kernel to include the 'nat' module.

 

[SamTzu]

Yup, the signs are clear. The spring is in the air and so is LXC.

 

[lastguru]

I would say that OpenVZ is well alive and kicking... With vSwap support, memory management became far far better and easier. And all critical stability bugs has been fixed. As for LXC, I think it still does not support quotas (I many be mistaken, though), which is a show-stopper. And don't forget that the main contributor to LXC is OpenVZ

 

[jekader]

let me also add a few thoughts on this topic.
Last week I wanted to bring up some container VMs on my home pc, and as LXC requires no custom kernels, I decided to give it a try. The results were: in the default debian kernel you can not limit memory and drive space. Also, even if I give the VM access to just one CPU, I see all of them in the guest system, despite the fact that I can use just one of them.

So I need to recompile the kernel and still I won't get all that openVZ allows me to do.

I beleive LXC is not even close to being production ready yet.

Also, here's a link to a post where an openVZ developer talks about LXC and how it is already used and even developed by the openVZ team:
http://translate.google.com/transla...//ru-openvz.livejournal.com/1970.html&act=url
(the post is in Russian, but the translation is mostly readable)

 

[tux]

We all saw what happened with OpenVZ support on Proxmox last year. I believe this is just beginning signs of the end of OpenVZ development.

What do you see? I see a lot of development in the OpenVZ Community and Kernel. Also i see fast Kernel Fixes, support of new kernel RHEL6, new tools, better performance, updated OS templates (very 3 months). Every day millions of containers based on OpenVZ or Virtuozzo running stable and doing their job over years. This technology used by named companies (http://www.parallels.com/de/products/pvc/customer-case-studies/) and thousands of administrators every day. And now you compare a good practice and well trialed solution with - with what? - with this http://lxc.sourceforge.net/index.php/news/ . Stop kidding - and stop telling LXC is alternative for OpenVZ - if you never use it in production with 1000 contrainers.

I would very much like to move to LXC from OpenVZ since LXC does not require modified kernel and it's memory management is way better (no more complicated beans on my coffee.) .

Bullshit. The problems with Java under OpenVZ is a problem of Java - not OpenVZ. And with one parameter you can solve it.

My money is on LXC.

Last update on lxc (0.7.4.1) was on 14.3.2011. Since one year no update no new features. Last update on OpenVZ was yesterday - and over 50 updates, fixes and patches since last July http://wiki.openvz.org/News/updates .

Fact is OpenVZ is not safe to use in a meaningful way anymore. Too many linux components require CAPABILITY="Security Hole/MySQL injection vulnerability" to get it to work on OpenVZ these days.

Why is OpenVZ unsafe? How? "MySQL injection vulnerability" in my vz.conf? What are you talking about?