UDS is fast approaching, lets get the ball rolling on some pre-UDS discussions

[Tom Haddon]

On 04/10/12 00:18, Clint Byrum wrote:
> With UDS in Copenhagen just around the corner, I'd like to start some
> conversations regarding various topics we will be discussing there. We
> should aim at boiling these down to a few (4 - 6) blueprints to be
> discussed in the usual UDS format.

[...]

As we're starting to deploy services in production with Juju there are
two particular things that it's becoming obvious would *really* make a
massive difference to us:

https://bugs.launchpad.net/juju/+bug/1049340: Juju doesn't support
multi-region
- This means we can't do true HA across DCs for services that require
read-write backends (most do, obviously).
- It also means we need twice as much hardware, because each region is
isolated from the other, so each one needs at a minimum two of each
service components to provide HA.
- Also means we can't have Openstack and MaaS talking to each other as
they're different environments

https://bugs.launchpad.net/juju/+bug/803042: The "bootstrap node" needs
high availability
- If Zookeeper data store is lost, state is lost, entire environment has
to be rebuilt

I know you're not looking for a list of bug reports here, but just
wanted to make sure if they needed discussion to move them forward, that
UDS might be the time to discuss them.

Thanks, Tom