Re: Stability/Suitability of 2.6 kernel vserver

On Wed, 18 Aug 2004 17:52:34 -0700 (PDT), Roderick A. Anderson
<raanders@xxxxxxx> wrote:
> I've been semi-following what's going on and see that vs for the 2.6
> kernel is up to an RC of 10 or so.
> 
> Question is;  How's everyone that is using testing it feel about using it
> in light weight production?
> 
> Currently we use vservers for several custom web applications (one vserver
> per application), hosting a couple of databases (PostgreSQL and MySQL),
> running some custome e-mail reports (in perl), and hosting a few domains
> that want features that Windows can't offer.
>   I am the primary user for every system except the one running the web
> sites so security between the vservers and their users isn't a big issue.
> Nor any kind of quotas (yet).
> 
> Should I plan on _using_ a 2.6 kernel version of VServers in November?
> 
> Rod

Rod,
Let me just say that I've been using the 2.4.20ctx-17 (now antique)
for some time in a scenario similar to what you described with out
incident. I know that doesn't address your question with the 2.6
kernel at all, I was just using it as a spring board into another
point.

Regarding PostgreSQL (and probably MySQL too) running in the vserver,
some of the fine-tuning steps to get a database server running well
for high performance applications won't work well in a vserver. 
Specifically, because PostgreSQL runs best when it's data is stored on
the outer portions of the disk and with it's WAL on a separate disk
and better yet, stripped across disks you will find that write
performance (inserts and updates) will not be ideal.

This is no fault of the virtualization software; As a matter of fact,
I believe that the linux-vserver projects introduces less overhead
than other options.  I think instead the problem is related to the
inherant abstraction of the disk system that the virtual server
provides.  By that I mean the virtual server sees a single large
filesystem.

I've spent some time thinking about how to improve write performance
for a specific virtual server and I think the only easy solution is to
create a high performance raid array (i.e. raid 1+0 (or is it 0+1?))
and either put all your virtual servers on it (mount it as /vservers)
or put a single vserver on it (mount it as /vservers/vs1).

All the other options I can think of are much more complex and I
haven't tried any of them. I chose to put my db server on a separate
box and connect to it through a private network.  This provides at
least a 3 fold improvement of write performance.

Of course, maybe performance isn't that important in your case; Most
of my vservers are merely there to logically separate administration
tasks and performance is not the goal.  If so, ignore my post.

I hope you have a nice day,

-- 
Matthew Nuzum           | Makers of "Elite Content Management System"
www.followers.net               | View samples of Elite CMS in action
matt@xxxxxxxxxxxxx              | http://www.followers.net/portfolio/
_______________________________________________
Vserver mailing list
Vserver@xxxxxxxxxxxxxxxxxxxxxx
http://list.linux-vserver.org/mailman/listinfo/vserver