Glenn Lagasse wrote: > * Karen Tung (Karen.Tung at Sun.COM) wrote: >> Glenn Lagasse wrote: >>> Hi Sanjay, >>> >>> * sanjay nadkarni (Laptop) (Sanjay.Nadkarni at Sun.COM) wrote: >>> >>>>> I agree with all of this. I think implementing options 1 and 3 gets us >>>>> a lot and option 3 covers the cases where option 2 would have been >>>>> useful. >>>>> >>>>> Thanks for the feedback! >>>>> >>>>> >>>> It is possible that I might have missed this aspect in the >>>> discussion, but having a user provide a VM or providing the option >>>> to create one, does not obviate the fact that bad set of VM options >>>> can be chosen. So is the concern that Joe (or Karen) brought up >>>> earlier being addressed with any of the non-default options ? >>>> >>> I haven't come up with a good way to verify that the VM configuration is >>> actually valid for whatever the user wants to Install inside the VM. >>> >> Is there a way to query the VM to see what parameters >> it has been configured with? This might be useful for validating >> whether a user configured VM is fundamentally OK. > > Yes, there is. > >>> The memory requirements will be dictated by whatever the install >>> mechanism requires at a minimum (at this point, that's going to be the >>> bootable AI ISO that we need to create). I think that whatever minimum >>> we come up with for that is likely to work fine both during installation >>> as well as post-installtion (though perhaps not optimally). Sizing the >>> virtual disk is a whole other matter. I don't see how we can come up >>> with a way to verify that what the user says he wants to install (again >>> from a bootable AI ISO) will fit in a given size of virtual disk. Not >>> to mention the other considerations like how is the appliance creator >>> going to make the resultant image available based on it's size. >>> >>> In the case of using a pre-configured VM, we can do a sanity check on >>> the memory requirement to make sure that the bootable AI ISO will have >>> enough memory to actually do the work it needs to, but other than that I >>> don't see how we can verify any other settings in any useful way. >>> >>> >>> I'm giving this more thought, the stuff we'll really care about are >>> memory, disk size and possibly networking (since we'll need that working >>> to do the install once booted in the VM). As I said, I think we can >>> come up with a minimum requirement for memory. I'm less sure how we can >>> constrain disk size and networking. >>> >>> Cheers, >>> >>> >> I don't have an answer for how much memory or hard disk is the minimally >> required. It all depends on the ISO. However, we can make some educated >> guesses given the information we have today, and we can adjust accordingly, >> like if we ever defined what's the minimal set of required packages for >> Open Solaris. >> That would be the smallest size we would allow then. > > Right, we'd have to do something like that (guess). If we knew what the > installed size was going to be for a given set of packages then we could > do better, but I'm not aware of any such mechanism. As for networking, > verifying that the settings will actually 'work' is another hard problem > that I don't think we can really verify (other than making sure that > 'something' is configured for networking). >
Does this support the value of a prototype? The more experience we have the more we will know what a valid VM configuration will look like. I think we will need to make sure if an error is encountered because of a poorly configured VM that our error reporting indicates as much as possible exactly what failed and why. This will enable us, via testing, to fine tune the validation. "I think" ;) Thoughts? Joe
