Call to arms: Tinderbox 3.0

[Joe Marcus Clarke]

Ion-Mihai Tetcu wrote:
> On Thu, 10 Jul 2008 08:41:50 +0200
> Erwin Lansing <erwin at freebsd.org> wrote:
> 
>> On Wed, Jul 09, 2008 at 10:36:46PM -0400, Joe Marcus Clarke wrote:
>>> The good news is since people like miwi are using it pretty
>>> extensively, I think 3.0 is in fairly good shape.  The tasks I see
>>> need which need to be done are as follows:
>>>
>>> * Check for documentation completeness, and fill in any holes.
>> I was planning on installing a new personal tinderbox anyway, so I'll
>> go the 3.0 route here and see what I'm missing for a scratch install.
>>> * Decide on how upgrades will be done.  Will they be automated like
>>> between 1.x and 2.0, or should we simply instruct users to do a
>>> wholesale replacement of 2.x with 3.0?  Left to me, I was just
>>> going to document what needs to happen to install 3.0 from scratch,
>>> and forget the upgrade.
>>>
>> Too bad it's not left to you :-)  
> 
> I don't feel like rebuilding all the PT again on QA Tindy, thank you
> very much :-)

Does that mean you guys are willing to step up to write the migration code?

The schema has radically changed in 3.0.  For starters, all column names 
are lower case.  All the data would need to be migrated from the 2.x 
schema to the 3.0 schema.  On top of that, the .env files will need to 
be migrated to the etc/env system.

> 
>> Serouisly though, on the one hand, I
>> think a scratch install of the shared tb3/tb4 tinderboxes might not
>> be a bad idea as some things can be done easier, on the other hand,
>> it'll require quite a lot of work with around 60 users, each with
>> their own portstree on 3 branches and 2 architectures.  Starting from
>> scratch might also give an opportunity to see if more can be shared
>> across the two boxes.
>>
>> Basic setup right now, portstrees are shared via NFS, all the rest is
>> running independent in two databases.  The trouble I had at the time
>> was how to reuse a build on two archs, which I finally gave up on and
>> went for independent installations and some hackish scripts to create
>> new users.  So any thoughts on how to ease user administration and
>> share as much information as possible between an i386, an amd64, and
>> a sparc64 is in the making, preferably from one webinterface/database.
>>
>> -erwin
>>
>> PS: There were rumours about hardware being available at ISC, so I
>> might have to do a scratch install anyway there.
> 
> In which case I'd very much like to run QA Tindy on it. I'm working on
> extending the testing to other non-pointy-hat tested configs and I
> could really-really use more horse power. :-)

This might also make a nice test bed for doing a better multi-node TB. 
This is something I've always lacked having only one machine to dedicate 
to running TB.

Joe

-- 
Joe Marcus Clarke
FreeBSD GNOME Team	::	gnome at FreeBSD.org
FreeNode / #freebsd-gnome
http://www.FreeBSD.org/gnome