[Gllug] Small clusters

Rich Walker rw at shadow.org.uk
Thu May 5 21:03:34 UTC 2005


John Hearns <john.hearns at streamline-computing.com> writes:

> On Thu, 2005-05-05 at 20:05 +0100, Rich Walker wrote:
>> Hi,
>> 
>> What do people think about small cluster software? Specifically, 2
>> machines, starting out running Debian, with plenty of disk space, used
>> as generic servers with NFS/SAMBA exports, such that each machine has
>> one copy of all the disk space, and failover works for NFS+SAMBA.
>> 
>> Mosix is probably not the answer...
> Rich, are you referring to a failover type cluster or a compute cluster?
>
> As a coincidence, in the office today we discussed a 'cluster on wheels'
> project. We're pricing an eight-node turnkey cluster, delivered working
> to your desk. Hopefully running off (n) 13 amp plugs.
> We'll do a two node if you like!


Heh - tempt me not, John! (Probably out of the budget)

> If you're looking at a database or applications cluster,
> drop me an email offline re. http://www.emicnetworks.com/
> I'd be very interested to hear if this is useful to you.
> May be useful to us to let you have some test time.

No, this is more a generic fileserver and "workstation".

> Commercial over...
> For a two-node failover cluster you should look at Linux-HA.
> http://www.linux-ha.org
> The shared storage is needed for  NFS/SAMBA.
> At the higher end, you need a SCSI array with dual ports.
> The alternatives would be a Firewire disk and Firewire hub,

This is the kind of thing I was thinking to avoid.

> the bargain alternative is to use DRDB - think of it as software RAID
> over a network. http://www.drbd.org

That I had looked at. I was thinking that someone would probably have
worked out a method involving, say, NBD export of disk volumes bound
into RAID-1, tho' DRDB says it is an alternative to doing exactly that... 

> Think of shared SCSI as to DRBD as hardware RAID is to software RAID.
> Cut your cloth to suit your budget.

Right - probably it is what I want.

The idea is to avoid extra hardware, weird hardware and so on -
basically, to do the equivalent of:

A hosts /home
B has an up-to-date copy of /home
A fails 
B can take over with no data-loss (possibly not automatically.)

and in the mean time, B can be used for Other Things.

cheers, Rich.



-- 
rich walker         |  Shadow Robot Company | rw at shadow.org.uk
technical director     251 Liverpool Road   |
need a Hand?           London  N1 1LX       | +UK 20 7700 2487
www.shadow.org.uk/products/newhand.shtml
-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug




More information about the GLLUG mailing list