Function getRepairFunc

pkg/healthchecker/health_checker_linux.go:57–83

getRepairFunc returns the repair function based on the component.

(hco *options.HealthCheckerOptions)

// getRepairFunc returns the repair function based on the component.
func getRepairFunc(hco *options.HealthCheckerOptions) func() {
	// Use `systemctl kill` instead of `systemctl restart` for the repair function.
	// We start to rely on the kernel message difference for the two commands to
	// indicate if the component restart is due to an administrative plan (restart)
	// or a system issue that needs repair (kill).
	// See https://github.com/kubernetes/node-problem-detector/issues/847.
	switch hco.Component {
	case types.DockerComponent:
		// Use "docker ps" for docker health check. Not using crictl for docker to remove
		// dependency on the kubelet.
		return func() {
			if _, err := execCommand(types.CmdTimeout, "pkill", "-SIGUSR1", "dockerd"); err != nil {
				klog.Errorf("Failed to send SIGUSR1 to dockerd: %v", err)
			}
			if _, err := execCommand(types.CmdTimeout, "systemctl", "kill", "--kill-who=main", hco.Service); err != nil {
				klog.Errorf("Failed to kill service %s: %v", hco.Service, err)
			}
		}
	default:
		// Just kill the service for all other components
		return func() {
			if _, err := execCommand(types.CmdTimeout, "systemctl", "kill", "--kill-who=main", hco.Service); err != nil {
				klog.Errorf("Failed to kill service %s: %v", hco.Service, err)
			}
		}
	}
}

// checkForPattern returns (true, nil) if logPattern occurs less than logCountThreshold number of times since last
// service restart. (false, nil) otherwise.