[Linux-HA] [right one] Please can anybody help: switch from v1 to v2 with 25 nodes

[Ragnar Kjørstad]

On Wed, Mar 14, 2007 at 01:32:31PM -0600, Alan Robertson wrote:
> > Why is the UUID not included e.g. in ha.cf? 
> > And why not in text format instead of binary?
> > And if the user didn't specify a UUID, why not generate one based on
> > uname (e.g. md5sum converted into UUID format) instead of using a random
> > UUID?
> > 
> >> but yes, extreme caution should be used when even contemplating
> >> looking at this file
> 
> I thought that was a little extreme ;-)
> 
> A UUID is actually an industry-standard data structure.  As a storage
> guy, Ragnar presumably knows all about them.  You can generate UUIDs
> from node names if you want (another config option).

Ah, I wasn't awere of that one. (But I'm glad I wasn't, because that
forced us to do the _right_ thing (generating hb_uuid)).
> 
> But, that's not what I wanted to do - because it doesn't allow you to
> change host names, nor to (in the future) merge clusters which _might_
> have conflicting names.

Yes, using UUIDs for this is a good idea.

And maybe the random UUIDs is better than the user having to specify the
UUIDs in ha.cf. So, never mind.


-- 
Ragnar Kjørstad
Software Engineer
Scali - http://www.scali.com
Scaling the Linux Datacenter