Open Stack

Michael Still <mikal at stillhq.com> wrote on 11/13/2013 04:27:17 PM:
> ...
> This was actually discussed in the session as an example of how other
> projects handle these problems. Our concerns (IIRC) were that it would
> take even more patches to land, and each of those patches is quite
> hard to land in nova these days.

More patches than what?  I thought the concern was that there is no 
solution, not a competition between two solutions.  Is there some simpler 
solution to the delete column case in which each migration can be undone 
after being partially applied?

> Additionally, it increases the
> complexity of our code a lot, because we have to handle databases in
> all possible states because of how our continuous deployment model
> works.

I'm not sure I understand this.  The approach I cited has code that always 
works, and each rev of the code needs to co-exist with only the previous 
rev and the following rev.

Thanks,
Mike
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20131114/5c1cfb22/attachment.html>