Ensemble shakes hands with MediaWiki

Ahmed Kamal kim0 at ubuntu.com
Fri Jul 8 16:25:55 UTC 2011

Yesterday I had contacted Ryan Lane, from the mediawiki Ops team. This 
is a follow up to the previous reach-out before the Dublin sprint. The 
goal was to get MW excited as an upstream to using Ensemble for their 
own operations. Ryan is building a private OpenStack cloud (which they 
refer to as "the cluster"). My understanding is that this cloud is to be 
used for operations testing and staging. Since I believe this is 
interesting to many, here is an overview of how the conversation progressed.

- Ryan had some misconceptions regarding Ensemble, namely that it is 
"good for initial deployment, but not long term management of systems"
- Ryan would like to "open" their ops formulas, just like they do MW
- Myself and Mark Mims addressed concern points for Ryan, namely that 
can drive not only clouds, but also (soon) LXC containers and physical 
hardware. That it operates at service level, not machine level. And that 
we're all about sharing
- We explained the concept of relations, events and how it makes 
everything easier and more consistent
- Ryan was interested in any direct benefits Ensemble would offer over 
puppet/chef. Also was interested regarding formula upgrades in a live 
- We explained how having intelligent services, that know how to react 
to events make managing and scaling services much simpler. The formula 
upgrades feature was mentioned to be supported. It could rapidly evolve 
- Ryan mentions their software is deployed separately from the OS
- We agree that the openstack cloud is a good ground for experimenting 
with next-gen devops tools
- Ryan agrees to let us have access to their openstack cloud, where the 
Ensemble community would attempt to model their environment in Ensemble.
- Ryan mentions having lots of ops documentation on 
http://wikitech.wikimedia.org and configuration files at 
- Ryan will be helping the Ensemble community with this project, and 
will try to attract ops volunteers to help write and maintain the formulas

Here's the log 

Next Steps: The way I see it, this can be a good step to have MW as the 
first upstream to adopt Ensemble in their internal ops. Should we want 
to proceed with this, we need someone to step up and engage with 
modeling the MW ops environment in Ensemble. Any takers for this ?

More information about the Ensemble mailing list