Wednesday, Stage, 16:00

• Convener / Facilitator
• Scott Russell

Discussion on using Terraforma, or chef metal. Pretty much most folks use their own orchestration frameworks that are home grown/sown together.

Chef recipes for infrastructure, is chef metal.

Introduced machine image, and has notifications between machines.

#Chef metal 1st public release in January. Approaching 1.0. Plugins for docker, AWS, rackers, digital ocean, fog, and Open stack. AWS released SNS and SQS, and amazon's ruby SDK. Idempotently run these again and again. Teardown to be written.

Matt Ray wrote spiceweasel.

What synchronisation mechanisms are people using, latches' etc. Specific example. deploying a HA cluster, HA proxy for load balancing, same service across multiple nodes. Or in active/passive environment,

CHef 12, push jobs coming, knife-ssh on crack, zeroQ based, and has a concept of quorums, Converge these nodes and then run this command. chef-zookeeper etc, are currently available,

look are recipe, 25,50 or 100 nodes, on AWS, Rackspace, then it will ask for all those nodes, and then register them with the chef server, does pre-allocation, and continues on, if they fail, they can be repeated, so they can be deleted. Chef server was never meant to be an orchestrator.

Metal has the concept of a provisioning node, so this can be a node that is created by chef, and then continues the chef metal task. Hoping for 1.0 soon. other devices are on the roadmap, eg, network devices, storage etc.

Expect someone to do a metal UI, spiceweasel 3 will probably have a yaml format. 4 network things, arista, cumulus, juniper have a chef agent, plus cisco9000.
Testkitchen and metal are converging on how they provide cloud drivers?

hanlon, available, mesos, kubernetes, to fleet, not yet resourced, but probably will be soon...