Multi Site Chef Server Availability - chef-boneyard/chef-summit-2014 GitHub Wiki

Location

Thursday, Medina, 10:30

Put all config policy into version control and use a synchronization process to push that data to your individual sites. (Basically, use automated release practices and version control for your big-scale production Chef changes)
- Note that this is scoped to one organization...
Write client handler that ships relevant node data to another Chef server (or replication service?).
Requires people to be explicit about the way they load data about their environments - because you lose indexing across servers, you can't run Chef search in data center 1 and get timely responses about nodes from data center 2.

Wide IP / central redirect service? (May require client modification to follow 302s)
DNS search suffix? Directional/Anycast DNS?

A big, messy problem. This is why Chef recommends one-server-per-data-center and using object-based replication.
Search remains a different problem domain from keeping Chef objects in sync.
- Add optional scope to Chef searches?
How do you replicate user public keys from one Chef server to another?
- If only OC-ID held this data instead - that'd be a much easier service to replicate...

Most of the people in the room haven't played with chef-sync (Chef 12 enterprise add-on) yet. (Doesn't solve the federated search issue, though)
ChefCorp could be a little more prescriptive (white papers? Or provide some models - cookbook(s)? chef-metal script that does this for us? ) with regards to Chef server deployments that satisfy some or all of these scenarios.

It is also concluded and resolved that Peter Capaldi is doing a great job as the Doctor.