Update the stackdriver agents yaml to include a deployment for cluster level resources

cluster/addons/metadata-agent/stackdriver/metadata-agent.yaml

@@ -7,6 +7,19 @@ metadata:
     kubernetes.io/cluster-service: "true"
     addonmanager.kubernetes.io/mode: Reconcile
 ---
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: metadata-agent-config
+  namespace: kube-system
+data:
+  node_level.conf: |-
+    KubernetesUseWatch: true
+    KubernetesClusterLevelMetadata: false
+  cluster_level.conf: |-
+    KubernetesUseWatch: true
+    KubernetesClusterLevelMetadata: true
+---
 kind: DaemonSet
 apiVersion: extensions/v1beta1
 metadata:
@@ -30,6 +43,24 @@ spec:
       - image: gcr.io/stackdriver-agents/stackdriver-metadata-agent:{{ metadata_agent_version }}
         imagePullPolicy: IfNotPresent
         name: metadata-agent
+        livenessProbe:
+          exec:
+            command:
+              - /bin/bash
+              - c
+              - |
+                if [[ -f /var/run/metadata-agent/health/unhealthy ]]; then
+                  exit 1;
+                fi
+          periodSeconds: 10
+          failureThreshold: 1
+          successThreshold: 1
+        volumeMounts:
+          - name: metadata-agent-config-volume
+            mountPath: /etc/config
+        command:
+          - /opt/stackdriver/metadata/sbin/metadatad
+          - --config-file=/etc/config/node_level.conf
         ports:
         - containerPort: 8000
           hostPort: 8799
@@ -42,7 +73,73 @@ spec:
       restartPolicy: Always
       schedulerName: default-scheduler
       terminationGracePeriodSeconds: 30
+      volumes:
+        - name: metadata-agent-config-volume
+          configMap:
+            name: metadata-agent-config
   updateStrategy:
     rollingUpdate:
       maxUnavailable: 1
     type: RollingUpdate
+---
+kind: Deployment
+apiVersion: apps/v1beta1
+metadata:
+  labels:
+    app: metadata-agent-cluster-level
+    kubernetes.io/cluster-service: "true"
+    addonmanager.kubernetes.io/mode: Reconcile
+  name: metadata-agent-cluster-level
+  namespace: kube-system
+spec:
+  replicas: 1
+  selector:
+    matchLabels:
+      app: metadata-agent-cluster-level
+  template:
+    metadata:
+      labels:
+        app: metadata-agent-cluster-level
+    spec:
+      serviceAccountName: metadata-agent
+      containers:
+      - image: gcr.io/stackdriver-agents/stackdriver-metadata-agent:{{ metadata_agent_version }}
+        imagePullPolicy: IfNotPresent
+        name: metadata-agent
+        livenessProbe:
+          exec:
+            command:
+              - /bin/bash
+              - c
+              - |
+                if [[ -f /var/run/metadata-agent/health/unhealthy ]]; then
+                  exit 1;
+                fi
+          periodSeconds: 10
+          failureThreshold: 1
+          successThreshold: 1
+        volumeMounts:
+          - name: metadata-agent-config-volume
+            mountPath: /etc/config
+        command:
+          - /opt/stackdriver/metadata/sbin/metadatad
+          - --config-file=/etc/config/cluster_level.conf
+        ports:
+        - containerPort: 8000
+          protocol: TCP
+        resources:
+          requests:
+            cpu: {{ metadata_agent_cluster_level_cpu_request }}
+            memory: {{ metadata_agent_cluster_level_memory_request }}
+      dnsPolicy: ClusterFirst
+      restartPolicy: Always
+      schedulerName: default-scheduler
+      terminationGracePeriodSeconds: 30
+      volumes:
+        - name: metadata-agent-config-volume
+          configMap:
+            name: metadata-agent-config
+  strategy:
+    rollingUpdate:
+      maxUnavailable: 1
+    type: RollingUpdate

cluster/gce/config-default.sh

@@ -158,7 +158,7 @@ ENABLE_METRICS_SERVER="${KUBE_ENABLE_METRICS_SERVER:-true}"
 ENABLE_METADATA_AGENT="${KUBE_ENABLE_METADATA_AGENT:-none}"
 
 # Version tag of metadata agent
-METADATA_AGENT_VERSION="${KUBE_METADATA_AGENT_VERSION:-0.2-0.0.16-1}"
+METADATA_AGENT_VERSION="${KUBE_METADATA_AGENT_VERSION:-0.2-0.0.19-1}"
 
 # One special node out of NUM_NODES would be created of this type if specified.
 # Useful for scheduling heapster in large clusters with nodes of small size.

cluster/gce/config-test.sh

@@ -152,7 +152,7 @@ ENABLE_METRICS_SERVER="${KUBE_ENABLE_METRICS_SERVER:-true}"
 ENABLE_METADATA_AGENT="${KUBE_ENABLE_METADATA_AGENT:-none}"
 
 # Version tag of metadata agent
-METADATA_AGENT_VERSION="${KUBE_METADATA_AGENT_VERSION:-0.2-0.0.16-1}"
+METADATA_AGENT_VERSION="${KUBE_METADATA_AGENT_VERSION:-0.2-0.0.19-1}"
 
 # One special node out of NUM_NODES would be created of this type if specified.
 # Useful for scheduling heapster in large clusters with nodes of small size.

cluster/gce/gci/configure-helper.sh

@@ -2156,11 +2156,15 @@ EOF
        [[ "${METADATA_AGENT_VERSION:-}" != "" ]]; then
       metadata_agent_cpu_request="${METADATA_AGENT_CPU_REQUEST:-40m}"
       metadata_agent_memory_request="${METADATA_AGENT_MEMORY_REQUEST:-50Mi}"
+      metadata_agent_cluster_level_cpu_request="${METADATA_AGENT_CLUSTER_LEVEL_CPU_REQUEST:-40m}"
+      metadata_agent_cluster_level_memory_request="${METADATA_AGENT_CLUSTER_LEVEL_MEMORY_REQUEST:-50Mi}"
       setup-addon-manifests "addons" "metadata-agent/stackdriver"
-      daemon_set_yaml="${dst_dir}/metadata-agent/stackdriver/metadata-agent.yaml"
+      metadata_agent_yaml="${dst_dir}/metadata-agent/stackdriver/metadata-agent.yaml"
-      sed -i -e "s@{{ metadata_agent_version }}@${METADATA_AGENT_VERSION}@g" "${daemon_set_yaml}"
+      sed -i -e "s@{{ metadata_agent_version }}@${METADATA_AGENT_VERSION}@g" "${metadata_agent_yaml}"
-      sed -i -e "s@{{ metadata_agent_cpu_request }}@${metadata_agent_cpu_request}@g" "${daemon_set_yaml}"
+      sed -i -e "s@{{ metadata_agent_cpu_request }}@${metadata_agent_cpu_request}@g" "${metadata_agent_yaml}"
-      sed -i -e "s@{{ metadata_agent_memory_request }}@${metadata_agent_memory_request}@g" "${daemon_set_yaml}"
+      sed -i -e "s@{{ metadata_agent_memory_request }}@${metadata_agent_memory_request}@g" "${metadata_agent_yaml}"
+      sed -i -e "s@{{ metadata_agent_cluster_level_cpu_request }}@${metadata_agent_cluster_level_cpu_request}@g" "${metadata_agent_yaml}"
+      sed -i -e "s@{{ metadata_agent_cluster_level_memory_request }}@${metadata_agent_cluster_level_memory_request}@g" "${metadata_agent_yaml}"
     fi
   fi
   if [[ "${ENABLE_METRICS_SERVER:-}" == "true" ]]; then

cluster/gce/util.sh

@@ -679,6 +679,10 @@ ENABLE_CLUSTER_MONITORING: $(yaml-quote ${ENABLE_CLUSTER_MONITORING:-none})
 ENABLE_METRICS_SERVER: $(yaml-quote ${ENABLE_METRICS_SERVER:-false})
 ENABLE_METADATA_AGENT: $(yaml-quote ${ENABLE_METADATA_AGENT:-none})
 METADATA_AGENT_VERSION: $(yaml-quote ${METADATA_AGENT_VERSION:-})
+METADATA_AGENT_CPU_REQUEST: $(yaml-quote ${METADATA_AGENT_CPU_REQUEST:-})
+METADATA_AGENT_MEMORY_REQUEST: $(yaml-quote ${METADATA_AGENT_MEMORY_REQUEST:-})
+METADATA_AGENT_CLUSTER_LEVEL_CPU_REQUEST: $(yaml-quote ${METADATA_AGENT_CLUSTER_LEVEL_CPU_REQUEST:-})
+METADATA_AGENT_CLUSTER_LEVEL_MEMORY_REQUEST: $(yaml-quote ${METADATA_AGENT_CLUSTER_LEVEL_MEMORY_REQUEST:-})
 DOCKER_REGISTRY_MIRROR_URL: $(yaml-quote ${DOCKER_REGISTRY_MIRROR_URL:-})
 ENABLE_L7_LOADBALANCING: $(yaml-quote ${ENABLE_L7_LOADBALANCING:-none})
 ENABLE_CLUSTER_LOGGING: $(yaml-quote ${ENABLE_CLUSTER_LOGGING:-false})