On Thu, 2013-03-14 at 17:42 -0700, Ian Main wrote: > On Thu, Mar 14, 2013 at 03:19:06PM -0700, David Lutterkort wrote: > > Hi Ian, > > > > On Thu, 2013-03-14 at 13:36 -0700, Ian Main wrote: > > > A possibly more overarching issue is hardware profiles. Hardware profiles > > > are rather different in ovirt as compared to most other providers I've > > > worked with. Generally they just define 2 profiles, neither of which > > > really define anything specific, instead the user is left to define how > > > much ram/cpus etc they want at instance creation. > > > > > > I have been thinking about it and I really think this is a very > > > different concept/usage pattern from most clouds which in turn makes > > > the API generally inconsistent. In fact, I think deltacloud should > > > define a set of specific hardware profiles and then map them to the > > > native versions as best as possible (user configurable maybe?). > > > > I don't like the idea of DC mapping HWP; there's too much application > > policy in that. But there's two pretty easy things we can do: > > > > * change the builtin HWP; I am pretty open to what good HWP for > > RHEV-M are > > * allow admins/whoever runs the DC server to define HWP by editing > > the config file for the rhev-m driver (in > > config/drivers/rhevm.yaml in git) > > > > Would either/both of these help make your life easier ? > > Yes that would fix it for sure. However if we require that admins have > to configure HWPs for rhevm in order to use heat then that will have to > be part of the steps for using it.. Not the end of the world but less > than ideal imo. > > Could you not define defaults in the config that you can then edit? > This is what I am basically advocating. That and they would all be the > same by default (same names, similar definitions accross providers).
Yes, that would be easy enough - we just need somebody to tell us what better default HWP should look like. David
