- Kubernetes v1.34: Of Wind & Will (O' WaW) | Kubernetes

Here are a few of the highlights:

• Dynamic Resource Allocation (now Stable) – it now has first class handling of GPUs, FPGAs, and accelerators. It also has better visibility, smarter scheduling, and the ability to share slices. This I think will be a big deal for ML in production, but even more exciting for home lab GPU configurations

• Per-container restart policy (Alpha) – restart a single container instead of rescheduling the whole pod. Less disruption, more efficiency. Great in prod, and handy in labs when you’re testing multi-service pods.

• Short-lived ServiceAccount tokens (Beta/Default) – replaces long-lived imagePullSecrets. More secure and less secret management overhead.

• Built-In Pod mTLS (Alpha) – native short-lived certs for workload-to-workload encryption, no sidecars needed. Moves Kubernetes toward zero-trust by default.

• KYAML (Alpha) – stricter, safer YAML that avoids those frustrating “why did this deploy break on whitespace” problems.

• OCI artifact volumes (Beta) – mount configs, models, or binaries from a registry without bloating images. Clean and simple.

• Graceful windows node shutdown (Beta) – Windows nodes finally respect termination grace periods.

From an SRE’s perspective, this is a release that is all about stability, observability, and security. From a home labber’s perspective, I think it helps with some of the friction with running Kubernetes in a home lab.

I’m personally excited about per-container restarts and the DRA graduation. Those are two that solve real pain points that I’ve seen in both production and lab environments. I am curious what you all think. Will you be testing 1.34 right away? Or will you be waiting until these features harden a bit?