Adds new service config for tf-operator (kubeflow#3408)

* Adds new service config for tf-operator Moves port config to tf-operator service instead of deployment Moves Prometheus config to service as well * Adds service file for test Removes enablePrometheus flag * Removes status field from tf-operator service ks test * Spec ports from string to number
ricardov1 · Jun 6, 2019 · 32e4190 · 32e4190
1 parent 8f4bade
commit 32e4190
Show file tree

Hide file tree

Showing 2 changed files with 65 additions and 24 deletions.
diff --git a/kubeflow/tf-training/tests/tf-job_test.jsonnet b/kubeflow/tf-training/tests/tf-job_test.jsonnet
@@ -110,21 +110,9 @@ std.assertEqual(
       namespace: "test-kf-001",
     },
     spec: {
-      ports: [
-        {
-          name: "monitoring-port",
-          port: "8443",
-          targetPort: "8443"
-        }
-      ],
       replicas: 1,
       template: {
         metadata: {
-          annotations: {
-            "prometheus.io/path": "/metrics",
-            "prometheus.io/port": "8443",
-            "prometheus.io/scrape": "true"
-          },
           labels: {
             name: "tf-job-operator",
           },
@@ -181,6 +169,39 @@ std.assertEqual(
   }
 ) &&
 
+std.assertEqual(
+  tfjobv1.tfJobService,
+  {
+    apiVersion: 'v1',
+    kind: 'Service',
+    metadata: {
+      annotations: {
+        'prometheus.io/scrape': 'true',
+        'prometheus.io/path': '/metrics',
+        'prometheus.io/port': '8443',
+      },
+      labels: {
+        app: 'tf-job-operator',
+      },
+      name: 'tf-job-operator',
+      namespace: 'test-kf-001',
+    },
+    spec: {
+      ports: [
+        {
+          name: 'monitoring-port',
+          port: 8443,
+          targetPort: 8443,
+        },
+      ],
+      selector: {
+        name: 'tf-job-operator',
+      },
+      type: 'ClusterIP',
+    },
+  }
+) &&
+
 std.assertEqual(
   tfjobv1.tfUiDeployment,
   {

diff --git a/kubeflow/tf-training/tf-job-operator.libsonnet b/kubeflow/tf-training/tf-job-operator.libsonnet
@@ -157,23 +157,11 @@
       },
       spec: {
         replicas: 1,
-        ports: [
-          {
-            name: "monitoring-port",
-            port: params.monitoringPort,
-            targetPort: params.monitoringPort,
-          },
-        ],
         template: {
           metadata: {
             labels: {
               name: "tf-job-operator",
             },
-            annotations: {
-              "prometheus.io/scrape": 'true',
-              "prometheus.io/path": "/metrics",
-              "prometheus.io/port": params.monitoringPort,
-            },
           },
           spec: {
             containers: [
@@ -194,6 +182,37 @@
     },
     tfJobDeployment:: tfJobDeployment,
 
+    local tfJobService = {
+      apiVersion: "v1",
+      kind: "Service",
+      metadata: {
+        labels: {
+          app: "tf-job-operator",
+        },
+        name: params.name,
+        namespace: params.namespace,
+        annotations: {
+          "prometheus.io/scrape": "true",
+          "prometheus.io/path": "/metrics",
+          "prometheus.io/port": params.monitoringPort,
+        },
+      },
+      spec: {
+        ports: [
+          {
+            name: "monitoring-port",
+            port: std.parseInt(params.monitoringPort),
+            targetPort: std.parseInt(params.monitoringPort)
+          }
+        ],
+        selector: {
+          name: "tf-job-operator",
+        },
+        type: "ClusterIP",
+      },
+    },  // tfJobService
+    tfJobService:: tfJobService,
+
     local tfConfigMap = {
       apiVersion: "v1",
       data: {
@@ -571,6 +590,7 @@
     all:: [
       self.tfJobCrd,
       self.tfJobDeployment,
+      self.tfJobService,
       self.tfConfigMap,
       self.tfServiceAccount,
       self.tfOperatorRole,