*   Passes sourcesReady from kubelet to the device plugin to avoid removing inactive pods during the grace period of kubelet restart.
    *   Extends the gpu_device_plugin e2e_node test to verify that scheduled pods can continue to run even after a device plugin deletion and kubelet restart.

// scheduling guarantees, and they will not be restarted when they exit or when a Pod is
// removed or restarted. The kubelet may evict a Pod if an ephemeral container causes the
// Pod to exceed its resource allocation.
//
// To add an ephemeral container, use the ephemeralcontainers subresource of an existing
// Pod. Ephemeral containers may not be removed or restarted.
message EphemeralContainer {

Na maioria (ou em todos) os casos, há uma opção simples para habilitar a execução do contêiner na inicialização e habilitar reinicializações em falhas. Por exemplo, no Docker, é a opção de linha de comando `--restart`.

Sem usar contêineres, fazer aplicativos executarem na inicialização e com reinicializações pode ser trabalhoso e difícil. Mas quando **trabalhando com contêineres** em muitos casos essa funcionalidade é incluída por padrão. ✨

  * Fix pods failing to start if they specify a file as a volume subPath to mount. ([#45623](https://github.com/kubernetes/kubernetes/pull/45623), [@wongma7](https://github.com/wongma7))

- Terminated Pods on a node will not be re-admitted on kubelet restart. This fixes the problem of Completed Pods awaiting for the finalizer marked as Failed after the kubelet restart. ([#127207](https://github.com/kubernetes/kubernetes/pull/127207), [@SergeyKanzhelev](https://github.com/SergeyKanzhelev)) [SIG Node and Testing]

// 'freeze' is 'false' would resume all S3 API calls again.
// NOTE: once a tenant is frozen either two things needs to
// happen before resuming normal operations.
//   - Server needs to be restarted 'mc admin service restart'
//   - 'freeze' should be set to 'false' for this call
//     to resume normal operations.
func (sys *NotificationSys) ServiceFreeze(ctx context.Context, freeze bool) []NotificationPeerErr {

- Fixes a regression on Kubelet restart and pod statuses. (#105560, @rphillips) [SIG Node and Testing]
- Fixes kubelet memory regression in 1.22 (#105452, @liggitt) [SIG Node]
- Fixes the kubelet's ability to restart static pods (#105075, @rphillips) [SIG Node and Testing]

* Fix a bug where kubelet would not retry pod sandbox creation when the restart policy of the pod is Never ([#79451](https://github.com/kubernetes/kubernetes/pull/79451), [@yujuhong](https://github.com/yujuhong))

In den meisten (oder allen) Fällen gibt es eine einfache Option, um die Ausführung des Containers beim Hochfahren und Neustarts bei Fehlern zu ermöglichen. In Docker ist es beispielsweise die Befehlszeilenoption `--restart`.

			time.Sleep(d)
		}
	}
	signal := serviceSignal(si)
	switch signal {
	case serviceRestart, serviceStop:
		dryRun := vars.Get("dry-run") == "true" // This is only supported for `restart/stop`

		waitingDisks := waitingDrivesNode()
		if len(waitingDisks) > 0 {
			buf, err := json.Marshal(waitingDisks)
			if err != nil {
				return np, grid.NewRemoteErr(err)
			}