Declarative chain to Kubernetes multi clusters for automation of HA workloads.
summary
GARR builds an automated pipeline from bare metal to federated Kubernetes using MAAS, juju bundles, OpenStack and Kubernetes so a declarative Juju config spins a host cluster plus two member clusters across three data centers. KubeFed propagates federated resources (template, placement, overrides) to create deployments with per-cluster replica counts (example: 2 replicas in cluster A, 3 in cluster B). Multi-cluster IngressDNS plus ExternalDNS translate endpoints into PowerDNS records. Current gaps: host-cluster single point of failure and multi-region persistent storage replication.
AI-written summary · about
chapters (16)
- 0:00Intro
- 0:28GARR Cloud architecture
- 1:57What we asked ourselves
- 2:08Our answer
- 3:36KubeFed actors
- 4:29Propagation
- 6:52Federated resources: example
- 9:31And now?
- 9:44ExternalDNS
- 10:15Multi-Cluster Ingress DNS
- 13:13Example IngressDNSRecord Object
- 14:12Example DNSEndpoint Object
- 14:42PowerDNS records
- 15:03Our architecture
- 15:58Future works
- 17:17Link Docs
related talks
46mCluster Federation in Kubernetes: Past, Present and the Future
34mKubernetes Multi-Cluster Operations without Federation
19mKeynote: Kubernetes Federation
33mMcK8s: Container Orchestration in Kubernetes Multi-Clusters
40mDeep Dive: Kubernetes SIG Multicluster
42mSIG Multi-Cluster Deep Dive
35mKubernetes journey
36mThe Architecture of a Multi-Cloud Environment with Kubernetes