-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Hi,
> I was wondering what was better (regarding performance)
between using
> disk images for domU and just "netbooting"
them from dom0 ?
I have no hard data on this as I haven't tested it. But
both "work
fine" for a setup with a dozen [not too heavily
utilized] VMs on a
single physical machine.
> My current configuration is:
> - every dom running NetBSD
> - system and pkgsrc in disk images (4Go)
> - data accessed by domUs from dom0 via NFS.
>
> My thinking configuration would be:
> - every doms running NetBSD
> - system, packages and data accessed by domUs from dom0
via NFS/
> netboot.
>
> any pros/cons ?
>
> Going further, as every domUs (or at least 99%) will be
NetBSD's, what
> about using a single NFS shared system/packages via NFS
with a
> layer for
> conf and temp files.
> Something like:
> domU1:
> dom0:/exports/domUs / nfs
> dom0:/exports/domU1 /mnt/specific nfs
> /mnt/specific / union
> domU2:
> dom0:/exports/domUs / nfs
> dom0:/exports/domU2 /mnt/specific nfs
> /mnt/specific / union
>
> binary updates are done (once) in dom0:/exports/domUs.
> each vmachine has it's own conf in
dom0:/exports/domU*.
> sounds crazy ?
Yes
I think you'll spend more effort fighting the standard
setup than it
is worth.
I ended up with sharing /usr/share + /usr/X11R6 + /usr/pkg
via NFS
and the rest in machine specific images. That brought down
the size
of the individual images to less than 200MB. But this is
because I
need to be able to install local software on each machine.
If that
isn't needed then you can probably share all of /usr and
shrink
images to ~50MB.
Local config stuff I manage something à la
for vm in $machines ; do
copy template.img $vm.img
mount $vm.img somewhere
rsync $vm specific stuff into somewhere
umount
done
Johan
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.0 (Darwin)
iD8DBQFETM3ZKJmr+nqSTbYRAiZvAKCegJOEsN+P+fnhPQJwMpeZINonuACe
Io92
aGwVVr+X44SdJsIrTT+RST0=
=WGnV
-----END PGP SIGNATURE-----
|