[clug] Clustering file systems, without excluding windows

Paul TBBle Hampson Paul.Hampson at Pobox.com
Thu Aug 16 16:31:45 GMT 2007

On Thu, Aug 16, 2007 at 02:54:05PM +1000, Robert Edwards wrote:
> Speaking of ATA over Ethernet (AoE), we are just in the process of
> commissioning a Coraid SR1521 15-disk AoE rack (with 8 x 750GB drives).

> I might be at a stage to lug it downstairs at next weeks CLUG meeting
> if anyone wants an up-close squiz at the innards and a short talk about
> how it all fits together.

> At this stage, I am looking at Lustre and GFS as two candidates for
> the clustered F/S. Andrew Tridgell spoke about IBMs proprietry GPFS
> during his recent CLUG talk on Samba Clustering - we aren't interested
> in going there at this stage. I have no interest in trying to make
> mine "work with Windoze", so can't help on that at all.

Heh. Went to reread up on Lustre, to work out why I took it off my list
the first time, and was reminded that clustered file systems need common
UID/GIDs between the servers. I've not set that up yet, so I guess
clustering's off my chart in the end, since I really don't want to put
that much time into LDAP.

On the other hand, the Lustre homepage mentions 'Lustre on DRBD', which
reminded me of the existence of DRBD (Duplicated Redundant Block Device
or something like that, for those playing along at home).

So now I'm thinking I'll just go with the traditional linux-ha failover
approach. That was my plan when I built these machines in the first
place, anyway.

Actually, a thought occurs to me here. If I were to Xen up each of these
machines, so they were really only running one domU per machine (for
now, consolidation is something else I should consider) I could then
throw a lot of hard disk in a slower machine, and make it DRDB-mirror
each of the other machines. Then if any one machine dies, it gets
booted on the mirror box, and continues like it had a server crash.

This means I only need one extra machine, with the risk that if I have a
multiple-server failure event (ie. mail server's rails snap, it drops
through the RADIUS server...) they both recover, just slower.

Of course, if I'm making each box its own Xen domU, I may as well throw
the SAN in the back there and score a live-migration setup for free.

I've always been tetering on the edge of just making every hosted
website it's own domU, although I can't do that for the windows-hosted
sites as I don't want to pay for the one license I am paying for, let
alone several dozen licenses. And of course I'd run out of IPs pretty
quick-smart. There's gotta be a better way to deal with that.

Reverse proxy bouncing to private IPs based on domain? Hell, reverse
proxy converting to IPv6, all domUs serve only only IPv6. Neither of
those solves for anything but www. Reverse FTP proxy based on login name
would be possible, but evil.

Dammit, this is the future. Rocket cars are all well and good, where's
my damn /48 allocation? (OK, I lie. I have a /48 for my home. But not
nearly enough people can _reach_ it.)

Thanks Bob, M0les. I've now turned one hard-to-wrap-my-head around
problem into a completely different but more fun for me to solve problem.

Well... "problemspace". ^_^

Paul "TBBle" Hampson, B.Sc, LPI, MCSE
Very-later-year Asian Studies student, ANU
The Boss, Bubblesworth Pty Ltd (ABN: 51 095 284 361)
Paul.Hampson at Pobox.com

Of course Pacman didn't influence us as kids. If it did,
we'd be running around in darkened rooms, popping pills and
listening to repetitive music.
 -- Kristian Wilson, Nintendo, Inc, 1989

License: http://creativecommons.org/licenses/by/2.1/au/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://lists.samba.org/archive/linux/attachments/20070817/5cb373c4/attachment.bin

More information about the linux mailing list