[Openhpi-devel] Re: Re: Simulator Plugin Design Documents

[Sean D. <se...@da...>]

On Wed, May 18, 2005 at 12:13:02PM -0500, Corey Minyard wrote:
<snip>
> It is true that individual platforms only cover part of what HPI can do,=
=20
> but I think you could create a simulated IPMI platform (or perhaps a set=
=20
> of them) that can cover all of HPI.  It seems like HPI capabilities are=
=20
> very close to a direct subset of IPMI capabilities.

Having helped a little on the BladeCenter simulator... an IPMI simulator
wouldn't be able to touch some of the things we were doing.  It doesn't mean
that an IPMI simulator isn't a good tool, it just wasn't the right tool for
that job.

> But it may still be simpler amd more effective to do a direct plugin. =20
> That's obviously for you to decide :-).

I think for best coverage we want something like the following:

Testing OpenHPI:
 * generic simulator where we can do things like for very strange edge cases
   as if the hardware was misbehaved.  All this code can be automated and
   made as part of dist check phase
 * note: generic simulator could also be used to toy with hardware modeling,
   but this is a secondary criteria
 * shooting for being able to test 100% of spec (i.e. do much better than
   the 40% NA all plugins are running now)

Testing IPMI Plugins:
 * ipmi simulator for testing plugins
 * note: the ipmi plugins are a bit light on automated tests (dist check)
   and it would be nice to change that

Testing BladeCenter Plugin:
 * blade center simulator (fakes out snmp calls) runs during dist check
 * note: this went in about a year ago, there are 70+ automated test
   programs that give us about 50% code coverage for the blade center plugin


The end game creates a framework where regressions can be very quickly found
and fixed.  We've had a few such issues in the past and I'd like to get
enough of the critical paths tested to ensure we don't end up doing that any
more.

	-Sean

--=20
__________________________________________________________________

Sean Dague                                       Mid-Hudson Valley
sean at dague dot net                            Linux Users Group
http://dague.net                                 http://mhvlug.org

There is no silver bullet.  Plus, werewolves make better neighbors
than zombies, and they tend to keep the vampire population down.
__________________________________________________________________