// MaxUnavailableStatefulSet feature. The field applies to all pods in the range 0 to
  // Replicas-1. That means if there is any unavailable pod in the range 0 to Replicas-1, it
  // will be counted towards MaxUnavailable.
  // +optional
  optional k8s.io.apimachinery.pkg.util.intstr.IntOrString maxUnavailable = 2;
}

// Scale represents a scaling request for a resource.
message Scale {

upon discussion in the [ztunnel hairpinning doc](https://docs.google.com/document/d/1uM1c3zzoehiijh1ZpZuJ1-SzuVVupenv8r5yuCaFshs/edit#heading=h.dwbqvwmg6ud3))

When a ztunnel receives traffic (authenticated or not) from a workload, it will forward that traffic to the Waypoint proxy **after** applying any `TRANSPORT` layer policies (i.e. `Authorization`s). Thus, if the destination workload has at least the equivalent of a `STRICT` `PeerAuthentication`, unauthenticated traffic will be rejected...

   ([#123341](https://github.com/kubernetes/kubernetes/pull/123341), [@neolit123](https://github.com/neolit123))
- kubectl `port-forward` over websockets (tunneling SPDY) can now be enabled using an `Alpha` feature flag environment variable: KUBECTL_PORT_FORWARD_WEBSOCKETS=true. The API Server being communicated to must *also* have an `Alpha` feature flag enabled: PortForwardWebsockets.

# This is the same as the official TensorFlow builds.
# See https://developer.nvidia.com/cuda-gpus#compute
# `compute_XY` enables PTX embedding in addition to SASS. PTX
# is forward compatible beyond the current compute capability major
# release while SASS is only forward compatible inside the current
# major release. Example: sm_80 kernels can run on sm_89 GPUs but
# not on sm_90 GPUs. compute_80 kernels though can also run on sm_90 GPUs.

It used custom types in its declarations instead of standard Python types, but it was still a huge step forward.

It also was one of the first frameworks to generate a custom schema declaring the whole API in JSON.

    dep ->> operation: Run dependency, e.g. DB session
    opt raise
        operation -->> dep: Raise HTTPException
        dep -->> handler: Auto forward exception
        handler -->> client: HTTP error response
        operation -->> dep: Raise other exception
        dep -->> handler: Auto forward exception
    end
    operation ->> client: Return response to client

) {
  fine("${queue.name} ${String.format("%-22s", message)}: ${task.name}")
}

/**
 * Returns a duration in the nearest whole-number units like "999 µs" or "  1 s ". This rounds 0.5
 * units away from 0 and 0.499 towards 0. The smallest unit this returns is "µs"; the largest unit
 * it returns is "s". For values in [-499..499] this returns "  0 µs".
 *
 * The returned string attempts to be column-aligned to 6 characters. For negative and large values

- `kubeadm`: now respects user provided `kubeconfig` during discovery process ([#113998](https://github.com/kubernetes/kubernetes/pull/113998), [@SataQiu](https://github.com/SataQiu))
- `kubectl port-forward` now exits with exit code 1 when remote connection is
  lost ([#114460](https://github.com/kubernetes/kubernetes/pull/114460), [@brianpursley](https://github.com/brianpursley))

* Pods that are marked for deletion and have exceeded their grace period, but are not yet deleted, no longer count toward the resource quota.([#46542](https://github.com/kubernetes/kubernetes/pull/46542))


#### Dynamic Admission Control

- `KMSv1` is deprecated and will only receive security updates going forward. Use `KMSv2` instead. In a future release, Set `--feature-gates=KMSv1=true` to use the deprecated KMSv1 feature. ([#119007](https://github.com/kubernetes/kubernetes/pull/119007), [@aramase](https://github.com/aramase))