chore: minor changes

Signed-off-by: Ajay Mishra <ajmishra@nvidia.com>

Author: XRFXLP

distros/kubernetes/nvsentinel/charts/janitor-provider/values.yaml

@@ -114,7 +114,9 @@ csp:
     # Container image used for the reboot Job
     rebootImage: "public.ecr.aws/docker/library/busybox:1.37.0"
     # Use the Linux Magic SysRq trigger for an immediate kernel reboot instead of
-    # running chroot /host reboot.
+    # running chroot /host reboot. Only enable this when target nodes support
+    # Linux Magic SysRq and their kernel/config permissions allow an immediate
+    # reboot through /proc/sysrq-trigger.
     useSysrqReboot: false
     # Namespace for reboot Jobs (defaults to the janitor-provider's namespace)
     rebootJobNamespace: ""

docs/configuration/README.md

@@ -104,6 +104,5 @@ Each module has additional configuration options documented in its dedicated gui
 - [Fault Quarantine](./fault-quarantine.md)
 - [Node Drainer](./node-drainer.md)
 - [Fault Remediation](./fault-remediation.md)
-- [Janitor Provider](./janitor-provider.md)
 - [Preflight](./preflight.md)
 - [Event Exporter](./event-exporter.md)

docs/designs/028-generic-baremetal-reboot-provider.md

@@ -33,7 +33,7 @@ flowchart TD
 
 ## Decision
 
-Add a **generic provider** (`CSP=generic`) that reboots nodes via a privileged Kubernetes Job. By default the Job runs `chroot /host /sbin/reboot`; deployments can opt into a Linux Magic SysRq reboot for environments where the normal reboot path wedges the node. This follows the Job-based pattern from GPU Reset ([ADR-019](019-janitor-gpu-reset.md)). It is a named provider in the factory switch, just like `aws`, `gcp`, or `kind`.
+Add a **generic provider** (`CSP=generic`) that reboots nodes via a privileged Kubernetes Job. By default the Job runs `chroot /host reboot`; deployments can opt into a Linux Magic SysRq reboot for environments where the normal reboot path wedges the node. This follows the Job-based pattern from GPU Reset ([ADR-019](019-janitor-gpu-reset.md)). It is a named provider in the factory switch, just like `aws`, `gcp`, or `kind`.
 
 ## Implementation
 
@@ -69,7 +69,7 @@ sequenceDiagram
     GP-->>JC: requestID = pre-reboot bootID
 
     K8s->>Job: Schedule pod on target node
-    Job->>Node: chroot /host /sbin/reboot or echo b > /proc/sysrq-trigger
+    Job->>Node: chroot /host reboot or echo b > /proc/sysrq-trigger
     Note over Node: Node reboots, pod is killed
 
     Note over Node: Node boots back up, new bootID assigned
@@ -116,7 +116,7 @@ spec:
       containers:
         - name: reboot
           image: busybox:1.37
-          command: ["chroot", "/host", "/sbin/reboot"]
+          command: ["chroot", "/host", "reboot"]
           securityContext:
             privileged: true
           volumeMounts:
@@ -138,7 +138,7 @@ spec:
 | `ttlSecondsAfterFinished` | `3600` | Auto-cleanup after 1h |
 | `tolerations` | `[{operator: Exists}]` | Target node is likely cordoned/tainted |
 | `restartPolicy` | `Never` | Do not restart after reboot |
-| Image | `busybox:1.37` | Only needs `chroot` and host `/sbin/reboot` |
+| Image | `busybox:1.37` | Only needs `chroot` and the host `reboot` command |
 
 #### IsNodeReady
 

janitor-provider/pkg/csp/generic/generic.go

@@ -156,35 +156,42 @@ func (c *Client) buildRebootJob(nodeName string) *batchv1.Job {
 	image := c.config.RebootImage
 	ttl := c.config.RebootJobTTL
 	command := c.config.rebootCommand()
-	volumeMounts := []corev1.VolumeMount{
-		{
-			Name:      "host-root",
-			MountPath: hostMountPath,
-		},
-	}
-	volumes := []corev1.Volume{
-		{
-			Name: "host-root",
-			VolumeSource: corev1.VolumeSource{
-				HostPath: &corev1.HostPathVolumeSource{
-					Path: "/",
+	var volumeMounts []corev1.VolumeMount
+	var volumes []corev1.Volume
+	if c.config.UseSysrqReboot {
+		volumeMounts = []corev1.VolumeMount{
+			{
+				Name:      "host-proc",
+				MountPath: hostProcMountPath,
+			},
+		}
+		volumes = []corev1.Volume{
+			{
+				Name: "host-proc",
+				VolumeSource: corev1.VolumeSource{
+					HostPath: &corev1.HostPathVolumeSource{
+						Path: "/proc",
+					},
 				},
 			},
-		},
-	}
-	if c.config.UseSysrqReboot {
-		volumeMounts = append(volumeMounts, corev1.VolumeMount{
-			Name:      "host-proc",
-			MountPath: hostProcMountPath,
-		})
-		volumes = append(volumes, corev1.Volume{
-			Name: "host-proc",
-			VolumeSource: corev1.VolumeSource{
-				HostPath: &corev1.HostPathVolumeSource{
-					Path: "/proc",
+		}
+	} else {
+		volumeMounts = []corev1.VolumeMount{
+			{
+				Name:      "host-root",
+				MountPath: hostMountPath,
+			},
+		}
+		volumes = []corev1.Volume{
+			{
+				Name: "host-root",
+				VolumeSource: corev1.VolumeSource{
+					HostPath: &corev1.HostPathVolumeSource{
+						Path: "/",
+					},
 				},
 			},
-		})
+		}
 	}
 
 	return &batchv1.Job{

janitor-provider/pkg/csp/generic/generic_test.go

@@ -116,18 +116,20 @@ func TestSendRebootSignal_CreatesSysrqJob(t *testing.T) {
 	container := job.Spec.Template.Spec.Containers[0]
 	assert.Equal(t, []string{"sh", "-c", "echo b > /host-proc/sysrq-trigger"}, container.Command)
 
-	assert.Contains(t, container.VolumeMounts, corev1.VolumeMount{
+	require.Len(t, container.VolumeMounts, 1)
+	assert.Equal(t, corev1.VolumeMount{
 		Name:      "host-proc",
 		MountPath: hostProcMountPath,
-	})
-	assert.Contains(t, job.Spec.Template.Spec.Volumes, corev1.Volume{
+	}, container.VolumeMounts[0])
+	require.Len(t, job.Spec.Template.Spec.Volumes, 1)
+	assert.Equal(t, corev1.Volume{
 		Name: "host-proc",
 		VolumeSource: corev1.VolumeSource{
 			HostPath: &corev1.HostPathVolumeSource{
 				Path: "/proc",
 			},
 		},
-	})
+	}, job.Spec.Template.Spec.Volumes[0])
 }
 
 func TestBuildRebootJob_UsesCommandRebootByDefault(t *testing.T) {