Hi Sarah,

* Sarah Jelinek (Sarah.Jelinek at Sun.COM) wrote:
> Hi All,
>
> Located at:
>
> http://opensolaris.org/os/project/caiman/CUD/Caiman-engine-redesign.pdf
>
> is a document that I would like folks to read prior to Wed meeting. It  
> includes the current Caiman high level design for the 'engine' pieces  
> and consumers of those pieces,and notes on areas to consider in this  
> redesign.

Here's my response to some of the information that was missing for VM
construction.

Question: 

-For the VM project, if we use DC we need to be able to instantiate and
configure the VM we are installing in. Again, likely a finalizer script but
need to understand the implications of the VM project

Response:

I've sent out a request soliciting people's opinion on what exactly our
requirements might be.  I'm not convinced that we'll want to own configuring
the VM from DC (at least any further than what is required to make the VM
boot using the ISO we create).

Question:

-Need requirements for this project with regard to: VM configuration, target
instantiation, target discovery, transfer of bits, post processing.

Response:

VM Configuration - Up in the air at this point, I know we'll need to do
'something' here but how much is up for debate.  Whatever it ends up being
this will likely just be a finalizer script that calls the VirtualBox cli
to set up the VM, configure it to boot the ISO we're going to start the
install from, start the VM and then poll the VM status to detect shutdown.
What is unclear is how the finalizer script configures the VM (see my
response to your question about instantiating and configuring a VM).

target instantiation - We'll definitely need to come up with something for
this to setup the virtual disk once we're booted inside the virtual machine.

target discovery - We may or may not need to do something here.  One possible
approach is that the manifest we use to perform the installation inside of
virtualbox specifies the name of the target disk to use.  Of course that
would require the 'user' to know what virtualbox is going to call the disk
when booted.  If we don't use that approach, then we'll need a mechanism that
I suspect looks similar to what AI does.

transfer of bits - currently, my design is based on issuing pkg commands
and installing from a network repository (specified in the installation
manifest).  This could be a finalizer script.  We might also leverage
libtransfer but I haven't looked at that.

post processing - It's not entirely clear what we'll need for this yet.
There will obviously be some things we'll have to do post transfer of
bits.  ICT should be useful in that regard.  There is also the post
processing that will need to take place after installation to convert
the VM into a portable OVF file.  But that's merely calling the
virtualbox cli to perform the necessary transform in a finalizer script.

Cheers,

-- 
Glenn

Reply via email to