Joseph J VLcek wrote: > Keith Mitchell wrote: >> >> >> Glenn Lagasse wrote: >>> * Joseph J VLcek (Joseph.Vlcek at Sun.COM) wrote: >>> >>>>>>> The requirement to shutdown the VM is really an ICT task which >>>>>>> >>>>>>>> you should add as part of your project. >>>>>>>> >>>>>>> I would not necessarily call this a strict requirement. It could >>>>>>> be done via ICT, or it could be done by having VirtualBox send >>>>>>> a shutdown signal to the machine, after determining from >>>>>>> progress reporting that the installation is complete. >>>>>>> >>>>> Ok, fair enough. We have to have VB send the shutdown signal >>>>> somehow, right? Is there a way to tell the VB VM to do this via >>>>> the api you will be using? >>>>> >>>> If the API can not support it it would not be difficult to shutdown >>>> the system once the install has completed. >>>> >>>> It could be accomplished a few different ways. >>>> >>>> - We could, as Sarah point out, have ICT do it. >>>> - We could have an SMF service built into the image which could do it. >>>> - bug 6556 "AI should provide an option for automatic reboot after >>>> an install" could be expanded to, in addition to reboot, provide >>>> for shutdown. >>>> >>> >>> This last option is what I was envisioning. The AI client already >>> provides for rebooting the client after installation. I merely want to >>> extend that to allow for a shutdown instead of reboot. >>> >>> While we *could* force the VM to shutdown by instructing virtualbox to >>> stop the VM, that's not ideal imo since it doesn't allow OpenSolaris to >>> perform a clean shutdown. >>> >>> >> You can have VirtualBox send an ACPI shutdown signal, which would >> cause the shutdown to occur gracefully. I agree that it's not an >> ideal solution though. > > Keith, How would VBox know when to send the ACPI? It would need to > have some why to know the install completed. Or am I missing something? > > Joe I was going along with the case of the guest OS indicating it was ready for shutdown, but (for some reason) unable to shutdown on its own. The point I was making is that it's possible to have a clean shutdown from outside the machine, so the AI image would not have a strict requirement to send the shutdown signal itself. The ability for AI to provide progress information is, however, a strict requirement.
-Keith > > >>> Sec 1.5 >>> >>>> >> Network access to an IPS repository (for generating the AI image or >>>> >> for installing the VM with) >>>> >>> > This should be "either access to an IPS repository, or an >>> installable > ISO image" I think. >>> >>> Yes and no. It depends on the design for the bootable AI image. If >>> that image contains everything you're going to install (as it does >>> today) then we only need access to an IPS repository when we're >>> constructing that image and not during the actual installation. If >>> however the AI image is just a 'skeleton' image that boots and >>> starts an >>> install but pulls the bits from a repo, then the VM will need access to >>> an IPS repo. I don't think I have a preference either way really. >>> Though there may be some issues with very large bootable AI images, but >>> I believe any known issues Dave already has a patch for (clofi issue >>> with archives greater than 4gb). >> Ah, right. I was seeing the AI .iso as having all the packages on it, >> but of course, that may not be the case. >> >> -Keith >
