2026-07-24
Cluster add-on upgrades, with Karpenter 1.14 and cert-manager 1.21
#add-on #kubernetes #update #upgrade #component #eks
We rolled out another round of cluster add-on upgrades. It is shipping to non-production clusters first, with production to follow in the coming week. This is a routine hygiene round: mostly patch releases, plus two minor version bumps (Karpenter and cert-manager) that need no action on your side.
What changed
- Karpenter 1.13 → 1.14. Adds support for AWS preview instance types and the new CapacityBuffers API, and fixes a Zonal Shift cache issue. No changes to your NodePools or provisioning behavior.
- cert-manager 1.20 → 1.21. Upstream renamed its metrics port and trimmed some default RBAC. Both are handled inside our chart configuration: certificates, issuers, monitoring and alerts all keep working unchanged.
- Monitoring stack refresh. A series of kube-prometheus-stack chart updates bundles Prometheus v3.13.1 and Grafana v12.7.2. No dashboards or alerts change for you.
Add-on upgrades
- alloy v1.17.1 (chart v1.10.1)
- amazon-eks-ami v20260714
- aws-load-balancer-controller v3.4.2
- aws-vpc-cni v1.22.3-eksbuild.1
- cert-manager v1.21.0
- Minor version bump. The controller’s metrics port was renamed upstream and some default RBAC was removed; both are absorbed by our chart configuration, so nothing changes for your certificates or monitoring.
- dcgm-exporter v4.8.3
- flux v2.9.2
- fluent-bit v5.0.9 (chart v0.57.9)
- karpenter v1.14.0
- Adds support for preview instance types and the CapacityBuffers API; fixes Zonal Shift cache hydration and secondary ENI IP accounting.
- kube-prometheus-stack v87.17.0 (Prometheus Operator v0.92.1)
- Bundles Prometheus v3.13.1 and Grafana v12.7.2. No dashboards or alerts change for you.
- grafana-loki v3.7.3 (chart v18.5.1 — chart-only update, Loki itself is unchanged)
- prometheus-blackbox-exporter v0.28.0 (chart v11.15.1 — chart-only update)
- tailscale-operator v1.98.9
- traefik v3.7.6 (chart v41.0.2)