Fully redundant LXD cluster
summary
Describes building a fully redundant 3-node cluster using Ceph for storage, OVN for networking, LXD for compute/management and BGP for external network redundancy, all on three identical second-hand servers hosted in a colo. Explains the hardware and design choices made to tolerate loss of a whole node or component failures, and the cost rationale versus cloud or rented dedicated servers.
AI-written summary · about
related talks
46mUnified Cluster Federation: A Hyperscaled Architecture of...
01h 41mBuild your own cloud with LXD
19mMicroCeph: Simplifying Storage
20mDeep Dive into Ceph Block Storage
39mMulti-clusters made easy with Liqo: getting rid of your clusters by keeping them all
39mKVM Modifications for Clustered Operation
37mBuilding the Multi-Cluster L7 Load Balancer
20mRDMA at Cloud Scale: The OCI Experience