Jail hot migration

[Klaus P. Ohrhallinger]

On Fri, Sep 10, 2010 at 10:28:09PM +0000, Andrew Hotlab wrote:

Hello;

> 
> I was really impressed by the live migration demo video and I wish to ask you a few questions about your solution...
> At this time VPS technology lacks resource limiting capability, do you think it could be compatible with the resource container project sponsored by the FreeBSD Foundation?

I heard about that resource container project, but don't know any details.
But it sounds like it could be adapted or integrated to my work.
I didn't put any effort in resource accounting/limiting yet because it's the
most boring part of the whole story, IMHO.

> Multiple virtual instances of FreeBSD running as VPS are required to share the same kernel, as in the Jail paradigm, or they may run different kernels, as in NetBSD (http://www.NetBSD.org/docs/rump/)?

They share the same kernel, altough it works different from jail.
Jail restricts processes from seeing and accessing certain resources,
while VPS duplicates almost any resource. Please see my paper for details.

> How many resources do you estimate in order to complete your project?

Well, if I get positive feedback at the EuroBSDCon and people want to have it
I can put quite a large amount of my time in it, but then it's still a long
way to go until production quality. Anyways I plan to publish patches and
binaries for testing in october.

> It seems that it will be far more difficult to implement the live migration feature on the Jail framework than in VPS, but how about offline migration and resource control? Some work has been done to achieve that with the current Jail paradigm, do you think that VPS will be able to do a better job in these areas too?

Live migration of jails is asking for some really nasty hacks. For example
imagine what to do if a process has a PID number that is already assigned
on the host you are migrating the jail to.

> Sorry if any question sounds stupid: I do not know much on the internals of the technologies I mentioned.
> Thank you very much for your work.
> Andrew
> P.S. I'm sorry for the format of the e-mail, I'm using this awful Hotmail web interface. :(
> 
>  		 	   		  

Best regards, Klaus
-- 
   Alle sagten ''es geht nicht''.
   Dann kam einer, der wusste das nicht.
   Der hat's einfach gemacht.