Spotlight on CoreOS

Transcript

1. 1.

   Spotlight on CoreOS martin@mbs3.org

2. 2.

   The Datacenter as a Computer - Focus on applications -

   Design for HW failure - No maintenance windows - Commodity hardware - Warehouse scale computer
3. 3.

   CoreOS - Micro Gentoo/Chrome (Kernel + Docker/LXC/nspawn/etc) - No package

   manager, just run in containers - Intended for HA clusters - Updates are applied automagically - SystemD (incl. managing containers)
4. 4.

   Etcd - Distributed key-value store - RESTful (HTTP + JSON)

   - Service discovery via announcements - Similar to Consul or Zookeeper
5. 5.

   Fleet - Cluster-level service management - Manage etcd and systemd

   as a single init sys - Schedule a specific number of containers, reschedule units on failure, placement rules
6. 6.

   Demos! - 3 node cluster with vagrant - Demo dillinger.io

   (online markdown editor) vagrant ssh core-01 fleetctl machine-list fleetctl list-units fleetctl submit dillinger.service fleetctl start dillinger.service fleetctl journal -follow dillinger.service https://github.com/coreos/coreos-vagrant http://lukebond.ghost.io/getting-started-with-coreos- and-docker-using-vagrant/