Sarah Jelinek wrote:
> Joseph J. VLcek wrote:
>> Glenn Lagasse wrote:
>>> Hey Keith,
>>>
>>> * Keith Mitchell (Keith.Mitchell at Sun.COM) wrote:
>>>  
>>>> Hi,
>>>>
>>>> I had a conversation with Sanjay a little bit ago. The subject came 
>>>> up  of how we would provide customization options that fall outside 
>>>> the  basic "install custom packages A, B and C to the VM" - 
>>>> configuration  options such as password, network settings, etc.
>>>>
>>>> Do we intend to support post install customizations? Would this be 
>>>> done  through SMF enhanced profiles, or in some other manner? Or 
>>>> are we just  going for the options we've discussed so far (package 
>>>> list, VM settings).
>>>>     
>>>
>>> That is an awfully good question.  My initial response is 'we'll 
>>> support
>>> whatever the installation engine we use supports'.  Which is vague I
>>> realize.  This is perhaps another requirement we might have for an AI
>>> client.  Right now, I can think of one option (which seem particularly
>>> unattractive to me).  The person constructing the images could create a
>>> new package which contains any customizations they want to introduce 
>>> and
>>> include that in the installation payload.  They would of course need to
>>> have some sort of SMF method to actually apply the changes (this is
>>> pretty much what we tell people who want to do scripting type 
>>> operations
>>> like they are used to in SVR4 land).
>>>
>>> This also depends on how the bootable AI image is designed.  For
>>> instance, if the AI image supports cpio transfer, then the deployer
>>> could create a finalizer script to customize the proto area of the AI
>>> image while it's being built to include whatever customizations he
>>> wants.  The problem there is that a) requiring people to write their 
>>> own
>>> finalizer scripts just seems very overweight and b) that's not really a
>>> supportable interface.  I do think that we might be able to say 'we'll
>>> use whatever the AI client provides in terms of customizing the
>>> installation payload'.  But then we'll need to express that requirement
>>> to the AI client redesign effort.
>>>
>>> Could you post a mail to caiman-discuss to solicit feedback on this
>>> problem and possible solutions people may have?  I'd like to hear
>>> people's thoughts on this.
>>>
>>> Thanks!
>>>
>>>   
>>
>> The Enhanced SMF project is also addressing this exact issue. I think 
>> until the eSMF project delivers we should just do what we can the way 
>> Glenn has described above. I don't think the VMC project should 
>> provide a "stop gap" solution to this until eSMF is available, at 
>> least not for our initial release. Perhaps we could document that 
>> this will be addressed in a later release.
>
> In terms of the VM project, customizations and eSMF.. certainly some 
> of the post installation system configuration could be done with this 
> when 'building' the VM image. However, this would require a reboot of 
> the VM and then the capture of the .ovf file, assuming this file 
> contains all of the configuration data upon export?
>
> At this point in time we do not have much in the way of configuration 
> specification that is allowed in the AI schema. This can be fixed, but 
> for the first release of the VMC project it isn't likely we will be 
> far enough along on the eSMF support to want to add this.

I agree.

>
> It isn't likely that all of it could be done via eSMF, however. We 
> need to understand what things are configurable in the VM with regard 
> to 'system' configuration, and what the VMC team is wanting to export 
> to the users.

OK

>
> thanks,
> sarah
> ****
>
>>
>> As Glenn suggests, I'm cross posting this to caiman-discuss to gather 
>> more thoughts...
>>
>> Joe
>>
>>
>>
>> _______________________________________________
>> caiman-discuss mailing list
>> caiman-discuss at opensolaris.org
>> http://mail.opensolaris.org/mailman/listinfo/caiman-discuss
>


Reply via email to