Never ever upgrade
summary
Reports show real outages from upgrade from 1.23 to 1.24 that fails in production because of DNS, and a systemd auto-upgrade that breaks monitoring. A sequential Kubernetes cluster rollup at a carmaker knocks out a region during lunch but recovers in two hours thanks to infra as code and failover. A 2.3 to 5.1 Elasticsearch reindex takes 7 months then shows performance regressions and weeks to rollback. Recommend blue-green replacement of entire clusters or DB replicas and move traffic gradually.
AI-written summary · about
related talks
46mEnd-Users Won't Even Notice, Painless Kubernetes Upgrades
19mUpgrading Your K8s Clusters: Key Considerations for a Smooth Transition
50mOps issues
25mAutomating Kubernetes Upgrades for Zero Downtime
29mElasticsearch 5 or and Bust
35mSafely Upgrading Kubernetes Clusters
32mWhen product meets real world data - how to analyze and tune deployments & upgrades
28mUpgrade Your Vault Using Failover