Re: [atomic-devel] Direct LVM by default

[Karanbir Singh <mail-lists karan org>]

Hi

On 09/23/2014 02:04 PM, Colin Walters wrote:
> Per previous discussion, I've been working on redoing the default Atomic
> setup to use direct LVM.  This is faster than loopback, and improves
> management ability.
> 
> See:
> https://bugzilla.redhat.com/show_bug.cgi?id=1145660
> https://github.com/projectatomic/rpm-ostree-toolbox/commit/0c7b373f6037d1f1d2ad2a46db0484721def44a8
> 
> Since we're still using -toolbox for CentOS, we should be able to
> upgrade there and spin a new cloud image from that.  For Fedora, I still
> need to land patches to the new compose tooling for this that uses
> Anaconda.
> 
> Some open items:
> 
> * How much space for the OS?  I chose 3G, given the host is 1G, leaving
> plenty for upgrades and host data.  But should we try for less?  More?

I think 3G is enough - the relative installed components dont really
evolve too much, apart from maybe the kernel.

> * How big should the disk be in general?  I chose 16G, reasoning that
> right now Docker containers are limited to 10G by default, so you can at
> least have a full one, plus other small ones.

8G disk is considered default/normal in most cloud vendor ecosystems,
and then layer in cloud-utils-growpart to consume whatever disk size is
allocatred to the instance. I wouldnt go too far away from 8G, it at all.

- KB

> 
> Of course, thanks to LVM all of this is mutable after the fact - you can
> attach other storage and migrate there.
> 
> 


-- 
Karanbir Singh
+44-207-0999389 | http://www.karan.org/ | twitter.com/kbsingh
GnuPG Key : http://www.karan.org/publickey.asc