Decentralized Live Migration Overview
summary
KubeVirt 1.6 adds Decentralized live migration that lets a running VM move across namespaces or clusters by creating a source migration and a receiver VMI synchronized over a synchronization controller via the migration network. Receiver VMIs use a wait-as-receiver runStrategy and a restoreRunStrategy annotation; source and target match with a migration ID and connection URL. Demo shows a cross-cluster migration of a timestamp-writing process with no downtime, using existing compute/storage live-migration mechanics plus state sync.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 0:45Live migration background
- 2:47Centralized migration process
- 4:40Need for decentralized approach
- 5:28Decentralized migration design
- 6:40Synchronization controller role
- 8:08Configuring receiver VMs
- 10:26Migration resource setup
- 12:04Migration demo
- 17:14Orchestration and conclusion
related talks
06mVM: When Compute Moves but Data Stays
18mAgility Unleashed: Migrating VMs to next generation open source platforms
24mOptimizing K8s and unlock blockers with Pod Live Migration
23mLive migration from VMware to OpenStack
18mIn-Memory Key-Value Store Live Migration with NetMigrate
31mVirtual What? Virtual Kubelet vs. Virtual Node
42mOperating a multi-tenancy Kubernetes service on KubeVirt
40mDeep Dive: Kubernetes (VMware) SIG
