Changes nvidia-gpu device plugin addon config settings:

- Runs as system critical pod - Makes resource limits to match its resource requets - Modifies test/e2e/scheduling/nvidia-gpus.go to cope with the recent change of running the device plugin as a system addon. - The resource settings of the addon is based on the test results from 8 nvidia-tesla-k80 gpus.

Changes nvidia-gpu device plugin addon config settings:
4a1a2051 · Jiaying Zhang · 849d7f85 · 4a1a2051 · 4a1a2051
Commit 4a1a2051 authored Nov 20, 2017 by Jiaying Zhang
Hide whitespace changes
Inline Side-by-side

Showing with 12 additions and 1 deletion

daemonset.yaml cluster/addons/device-plugins/nvidia-gpu/daemonset.yaml +7 -1

nvidia-gpus.go test/e2e/scheduling/nvidia-gpus.go +5 -0

No files found.
--- a/cluster/addons/device-plugins/nvidia-gpu/daemonset.yaml
+++ b/cluster/addons/device-plugins/nvidia-gpu/daemonset.yaml
@@ -11,7 +11,10 @@ spec:
    metadata:
      labels:
        k8s-app: nvidia-gpu-device-plugin
+      annotations:
+        scheduler.alpha.kubernetes.io/critical-pod: ''
    spec:
+      priorityClassName: system-node-critical
      affinity:
        nodeAffinity:
          requiredDuringSchedulingIgnoredDuringExecution:
@@ -34,7 +37,10 @@ spec:
        name: nvidia-gpu-device-plugin
        resources:
          requests:
-            cpu: 10m
+            cpu: 50m
+            memory: 10Mi
+          limits:
+            cpu: 50m
            memory: 10Mi
        securityContext:
          privileged: true

--- a/test/e2e/scheduling/nvidia-gpus.go
+++ b/test/e2e/scheduling/nvidia-gpus.go
@@ -183,6 +183,11 @@ func testNvidiaGPUsOnCOS(f *framework.Framework) {
 	pods, err := framework.WaitForControlledPods(f.ClientSet, ds.Namespace, ds.Name, extensionsinternal.Kind("DaemonSet"))
 	framework.ExpectNoError(err, "getting pods controlled by the daemonset")
+	devicepluginPods, err := framework.WaitForControlledPods(f.ClientSet, "kube-system", "nvidia-gpu-device-plugin", extensionsinternal.Kind("DaemonSet"))
+	if err == nil {
+		framework.Logf("Adding deviceplugin addon pod.")
+		pods.Items = append(pods.Items, devicepluginPods.Items...)
+	}
 	framework.Logf("Starting ResourceUsageGather for the created DaemonSet pods.")
 	rsgather, err := framework.NewResourceUsageGatherer(f.ClientSet, framework.ResourceGathererOptions{false, false, 2 * time.Second, 2 * time.Second, true}, pods)
 	framework.ExpectNoError(err, "creating ResourceUsageGather for the daemonset pods")