Paris OpenStack Summit - The Big Tent Casts a Long Shadow

So it's been a couple of weeks since I got back from the Paris Summit, but it's been a busy time and I haven't had time to write up all of my thoughts about it. Here's my belated attempt to do that.

From a purely personal perspective, I felt like this summit was much more useful than Atlanta. That's not a commentary on the summit itself, but more on my involvement in it. I don't know if I was just more comfortable jumping into discussions or if an extra six months working on everything provided me with stronger opinions, but regardless I felt like I contributed a lot more this time around. Some highlights of the discussions I was in:

Logging

Lots of good discussion around improving logging this cycle, which, as a TripleO contributor who is constantly frustrated by our useless logging at anything but DEBUG level, makes me very happy. We have a spec that is very close to ready IMHO, and once that merges we can get to work really improving things in this area.

Upgrades

Also lots of talk about upgrades. Since this and logging were the top two operator issues with OpenStack (by a long shot, IIRC), it's good that the focus is appropriately strong on them. I got the impression from some of the discussions that there's still a lot that isn't fully understood about how Nova is moving forward with their live upgrade mechanisms (and thus how the other projects can follow suit), but once all the mechanisms are in place it becomes more of an education issue than an implementation one, which should be an easier problem.

There's a long way to go, but I'm seeing a glimmer of light at the end of the live upgrades tunnel.

TripleO Review Policy

This was a long discussion, but I think it was valuable and the end result was a streamlining of the review process in TripleO. If this works out I could see it being adopted across many more projects (some have already expressed interest). Since core reviewer bandwidth is one of the biggest bottlenecks in OpenStack, optimizing it is important.

See the new review policy spec for a detailed discussion of the motivations and conclusions from this discussion.

Puppet/Chef/Ansible/etc.

While I will admit to having been poorly prepared for these discussions, it felt like there was a good dialog between some experts in the various tools and the TripleO team who are trying to implement their usage. I learned a lot and I think a general path forward was determined. It's early days for that work though, so we'll see where it goes from here.

Lunch

Not a technical point, but I thought some kudos were deserved here. Lunches at Le Meridien were terrific. The lines were kind of excessive the last two days, but I suspect that was a reflection of the quality and the fact that fewer people felt the need to go elsewhere to find food they wanted to eat.

There were a few lowlights too though:

Big Tent

The Big Tent sessions didn't feel like they made any significant progress toward an answer, at least from where I was sitting. Based on that, my suspicion is that a massive, all-at-once change to a different governance model just isn't going to happen. Hopefully we can find some incremental improvements to make along the way, and maybe some day we'll end up in a place everyone is happier with. This topic's probably above my pay grade though.

Containers

Based solely on this session, containers as a service may never be a thing in OpenStack. :-(

"The sky is falling" pessimism aside though, my reaction after the session concluded was less negative than while I was in the session. Honestly, I think the disconnect there (which was never explicitly called out in the session, to my knowledge) is that the folks working on containers want to collaborate with the OpenStack Compute team to make their new project fit in as well as possible, but the OpenStack Compute team is overloaded as it is and can't give them the time and effort they're looking for. This has led to a lot of wheel spinning in the effort to get the project off the ground.

I suspect the outcome of this topic is going to be that the containers team goes off and does their thing, and eventually they live as a largely separate entity under the Compute umbrella. But my crystal ball has been wrong before. :-)

In the end, I think it was a productive summit and we seem to be working on the right things to make OpenStack better for the people using it. I could probably write for another $TOO_MANY hours about everything that went on, but I have an RDO release for TripleO to work on so I'll stop here.