What's New in minikube v1.39.0: containerd Default, Kubernetes 1.37, Traefik Addon

💡 Quick Answer: minikube v1.39.0 makes containerd the default container runtime, adds support for Kubernetes v1.36/v1.37 (default now v1.37.0), ships a new Traefik Helm-based ingress addon, and brings Apple Silicon support to the VirtualBox driver. Existing clusters created with docker or cri-dockerd runtimes keep working, but any new minikube start without an explicit --container-runtime flag will now provision containerd.

The Problem

Operators tracking local-cluster tooling need to know whether upgrading minikube will change the behavior of existing scripts, CI jobs, or Kubernetes version pinning. v1.39.0 includes a default-runtime change and a cri-dockerd support reduction that can silently alter cluster provisioning if teams rely on implicit defaults.

The Solution

Upgrade the minikube binary, review the breaking changes below, and explicitly pin --container-runtime and --kubernetes-version in automation if the previous defaults (Docker runtime, older K8s version) must be preserved.

What’s New

Highlights

• Kubernetes v1.36 and v1.37 support added; default cluster version is now v1.37.0 (#23567, #23596)
• Default container runtime changed from Docker to containerd (#23562)
• New Traefik Helm-based ingress addon (#23225)
• VirtualBox driver now supports Apple Silicon with VirtualBox 7.1+ (#22861)

Other notable features

• Helm added to the minikube ISO, guest Helm client upgraded to Helm 4 (#23420, #23383)
• Configurable DNS servers for VM drivers (#23069)
• New --mdns flag to enable mDNS on VM start (#22964, #23208)
• Deterministic MAC addresses for drivers (#23172)
• vfkit auto-selects vmnet-shared network when available (#23040)
• VM drivers now start in parallel for faster minikube start (#23112)
• Preload images now fetched from GitHub by default, GCS as failover (#23087)

Notable bug fixes

• kubectl exec with cri-dockerd fixed on Kubernetes v1.36+ by disabling ExtendWebSocketsToKubelet (#22921)
• --extra-config=etcd.* no longer produces invalid kubeadm config (#23051)
• SSHRunner exit codes now correctly propagate instead of always returning 0 (#23364)
• image load no longer fails silently when the image is in use by a running container (#23034)

Breaking changes

• Default container runtime changed to containerd (#23562) — scripts assuming Docker (e.g. calling docker inside the node, or relying on minikube docker-env) will break unless they pass --container-runtime=docker explicitly.
• cri-dockerd support is now limited to x86 and arm64, and is migrated to pre-compiled GitHub binaries (#22922) — other architectures lose cri-dockerd support.

Upgrade Steps

Upgrade the minikube binary (Linux example; adapt for macOS/Windows per the official install docs):

curl -LO https://storage.googleapis.com/minikube/releases/v1.39.0/minikube-linux-amd64
sudo install minikube-linux-amd64 /usr/local/bin/minikube
minikube version

If existing automation depends on the Docker runtime or an older default Kubernetes version, delete and recreate the cluster with explicit flags rather than relying on defaults:

minikube delete
minikube start \
  --kubernetes-version=v1.37.0 \
  --container-runtime=containerd \
  --driver=docker

To keep the previous Docker runtime instead of adopting containerd:

minikube start --container-runtime=docker

Enable the new Traefik ingress addon if needed:

minikube addons enable traefik

Verify

minikube version
# confirm: minikube version: v1.39.0

kubectl version --short
# confirm server version matches the pinned/default v1.37.0

minikube ssh -- "sudo crictl info | grep -i containerd"
# confirm containerd is the active runtime (if using the new default)

kubectl get pods -A
# cluster-critical pods should be Running

Common Issues

• CI pipelines break after upgrade: caused by the containerd default change. Pin --container-runtime=docker explicitly in CI minikube start invocations until scripts are updated.
• cri-dockerd unavailable on non-x86/arm64 hosts: expected per the breaking change in #22922; switch to containerd or Docker runtime on those architectures.
• minikube docker-env behaves unexpectedly: this command is only meaningful with the Docker runtime; with containerd as default it no longer applies — use minikube image load workflows instead.

Best Practices

• Pin both --kubernetes-version and --container-runtime in scripts and CI to avoid surprises from future default changes.
• Test the new containerd default in a disposable cluster (minikube start -p test-containerd) before rolling the upgrade out to shared dev environments.
• Review the full v1.39.0 release notes for the complete list of ISO and driver-level changes not covered here.