The Leapfrog Upgrade Playbook
summary
Walmart leapfrogs 11 Kubernetes releases by provisioning a fresh leapfrog to fresh 1.32 cluster, backing up the old control plane (etcd ≈7GB, millions of objects), pruning temp namespaces, and performing a selective restore with parallel shards (shard by namespace or labels). Key operational changes: overclock open‑source backup/restore clients, tune API server, admission webhooks and Go GC, and temporarily disable APF; result cuts restore from ~12 hours to about 1 hour and keeps a warm rollback for quick flipback.
AI-written summary · about
chapters (12)
- 0:00Introduction to WCNP
- 2:47GitOps and platform architecture
- 3:44The challenge of leapfrogging
- 5:11Choosing the leapfrog strategy
- 8:06Execution of the failover
- 11:13Managing API deprecations
- 13:10Optimizing restore performance
- 15:11Tuning the stack components
- 18:45Parallelization and sharding
- 21:42Selective resource restoration
- 22:26Results and performance wins
- 22:56Best practices and conclusion
related talks
46mEnd-Users Won't Even Notice, Painless Kubernetes Upgrades
28mUpgrade Your Vault Using Failover
40mContinuously Deliver your Kubernetes Infrastructure
39mStrangling the Legacy with Kubernetes
29mChanging the Engine Mid-Flight: Zero-Downtime Ceph Upgrades
35mLeapp: Red Hat tool to upgrade from RHEL7 to RHEL8
25mFrom Duct Tape To Declarative: Playtika's Platform Overhaul at Unicorn Scale
33mUnleashing the Gorillacorn - DevOps in Very Large Enterprises