
--- ## Legacy - Deb Packages for openstack components - Each service on KVM -- ### Implication - Frozen repositories (sic) - Custom VM hypervisor... -- ### First step - Vendorization - All-In-One packages: - code - deps - libs --- ## Let's talk about boring metal "Classical" distributions are for pets (they have a name, you know their mood) By ops for ops -- ## Chromium OS / Chrome OS / Chromebooks - Upgrade in 1 shot - System in Read Only -- - Less divergence - More fiability --- ## CoreOS - systemd - container runner (rkt / docker) - etcd --- ## Containers - Server-centric - Application-centric --- ## Docker - Golang - Daemon - RESTful API - Image format - And more Registry, COW, orchestrator, SDN... -- ### Docker + OpenStack (on deploy side) Kolla - docker images - ansible playbook - wrapper env -> openstack conf --
--- ## Fleet
-- - sheduling - hardware fault recovery --
--- ## Kubernetes <3
-- - Service Discovery - Supervision - Only etcd as stateful component -- ### OpenStack over k8s - define your service (and way to access port / dns / IP...) - create images who are using it -- ### Stateful components - outside of cluster - using new-gen proxy - link k8s in/out with software cluster --- ## Final POC State - lightweight bootstrapper - 3 servers - fleet cluster OK - k8s cluster OK + HTTP gateway + DNS discovery - simple keystone - simple glance -- From scratch with 20 man/days! --- ## Difficulties -- OpenStack devs... Y NO 12factor!!! -- You can lead a horse to water but you can’t make him drink ---
--- Questions?