[OCF] Implementations of OCF resource agent API

[Lars Marowsky-Bree]

On 2004-10-02T15:51:00,
   Alan Robertson <alanr at unix.sh> said:

> >There are some trivial extensions rgmanager uses as part of the
> >metadata; but because it is XML, other RMs can simply ignore them.  Some
> >of the extensions are probably useful for the OCF RA API, but we'll see
> >how it works out.
> >
> >The most obvious (useful, I think) extension is shortdesc and longdesc
> >fields as a child of the resource-agent tag instead of just the
> >parameters ;)
> 
> I'm sure LMB would be happy to discuss this stuff with you.  I doubt that 
> very many people care about the subtleties of the XML representation.
> 
> This is not to say that they're unimportant, in fact agreeing on them is 
> quite important.
> 
> Can you explain more why you think this change should be made?

As it happens, Lon and I already had discussed these. The example
IPaddr2 agent in the heartbeat CVS actually already has these ;-)

They are needed to not only provide descriptions for the parameters one
by one, but also for the agent as a whole (doh). That they've not been
in from day 1 can be called an oversight I guess.

We'll need to make some extensions to the OCF RA API anyway for the
extended resource model as the implementation has shown so far, but they
are mostly some odd attributes here and there, moving the path and so
on, nothing truely substantial, just the regular fall-out which happens
when one actually goes and implements a spec ;-)

I've got that on my radar to feedback to the list here by the end of
November or so, when the implementation of the next generation heartbeat
Resource Manager has stabilized far enough for that and to make
qualified comments.


Sincerely,
    Lars Marowsky-Brée <lmb at suse.de>

-- 
High Availability & Clustering
SUSE Labs, Research and Development
SUSE LINUX AG - A Novell company