Remove TensorBoard related code in operator (#391)

Ref #347 Blocked until CI is running again. PS: Dashboard code is not changed. Signed-off-by: Ce Gao ce.gao@outlook.com
kubeflow · Feb 27, 2018 · 0759f7a · 0759f7a
1 parent a814670
commit 0759f7a
Show file tree

Hide file tree

Showing 23 changed files with 12 additions and 928 deletions.
diff --git a/README.md b/README.md
@@ -158,108 +158,6 @@ spec:
 Follow TensorFlow's [instructions](https://www.kubeflow.org/tutorials/using_gpu)
 for using GPUs.
 
-### Requesting a TensorBoard instance
-
-You can also ask the `TFJob` operator to create a TensorBoard instance
-by including a [TensorBoardSpec](https://github.com/kubeflow/tf-operator/blob/master/pkg/apis/tensorflow/v1alpha1/types.go#L95)
-in your job. The table below describes the important fields in
-[TensorBoardSpec](https://github.com/kubeflow/tf-operator/blob/master/pkg/apis/tensorflow/v1alpha1/types.go#L95).
-
-| Name | Description | Required | Default |
-|---|---|---|---|
-| `logDir` | Specifies the directory where TensorBoard will look to find TensorFlow event files that it can display | Yes | `None` |
-| `volumes` | `Volumes` information that will be passed to the TensorBoard `deployment` | No | [] |
-| `volumeMounts` | `VolumeMounts` information that will be passed to the TensorBoard `deployment` | No | [] |
-| `serviceType` | `ServiceType` information that will be passed to the TensorBoard `service`| No | `ClusterIP` |
-
-#### TensorBoard on Azure
-
-On Azure you can store your event files on an Azure Files and use
-volumes to make them available to TensorBoard.
-
-```
-apiVersion: "kubeflow.org/v1alpha1"
-kind: "TFJob"
-metadata:
-  name: "tf-smoke-gpu"
-spec:
-  replica_specs:
-    - replicas: 1
-      tfReplicaType: MASTER
-      template:
-        spec:
-          containers:
-            - image: gcr.io/tf-on-k8s-dogfood/tf_sample_gpu:latest
-              name: tensorflow
-              resources:
-                limits:
-                  alpha.kubernetes.io/nvidia-gpu: 1
-          restartPolicy: OnFailure
-  tensorboard:
-    logDir: /tmp/tensorflow
-    volumes:
-      - name: azurefile
-        azureFile:
-            secretName: azure-secret
-            shareName: data
-            readOnly: false
-    volumeMounts:
-      - mountPath: /tmp/tensorflow
-        name: azurefile
-
-```
-
-#### TensorBoard on GKE
-
-On GKE you can store your event files on GCS and TensorBoard/TensorFlow
-can read/write directly to GCS.
-
-```
-apiVersion: "kubeflow.org/v1alpha1"
-kind: "TFJob"
-metadata:
-  name: "tf-smoke-gpu"
-spec:
-  replica_specs:
-    - replicas: 1
-      tfPort: 2222
-      tfReplicaType: MASTER
-      template:
-        spec:
-          containers:
-            - image: gcr.io/tf-on-k8s-dogfood/tf_sample_gpu:latest
-              name: tensorflow
-              args:
-                - --log_dir=gs://my-bucket/logdir
-              resources:
-                limits:
-                  alpha.kubernetes.io/nvidia-gpu: 1
-          restartPolicy: OnFailure
-  tensorboard:
-    logDir: gs://my-bucket/logdir
-
-```
-
-#### Connecting to TensorBoard
-
-The TFJob operator will create a service named
-**tensorboard-$RUNTIME_ID** for your job. You can connect to it
-using the Kubernetes API Server proxy as follows
-
-Start the K8s proxy
-```
-kubectl proxy
-```
-
-In a web-browser open up
-
-```
-http://${PROXY}:8001/api/v1/proxy/namespaces/default/services/tensorboard-${RUNTIME_ID}:80/
-```
-
-Depending on how you configure the service for TensorBoard and cluster
-you can make TensorBoard available without using the K8s proxy.
-
 ## Monitoring your job
 
 To get the status of your job
@@ -324,11 +222,6 @@ spec:
         restartPolicy: OnFailure
     tfPort: 2222
     tfReplicaType: PS
-  tensorboard:
-    logDir: /tmp/tensorflow
-    serviceType: ""
-    volumeMounts: null
-    volumes: null
   tfImage: tensorflow/tensorflow:1.3.0
 status:
   conditions: null

diff --git a/dashboard/backend/handler/api_handler.go b/dashboard/backend/handler/api_handler.go
@@ -24,11 +24,10 @@ type APIHandler struct {
 }
 
 // TFJobDetail describe the specification of a TFJob
-// as well as related TensorBoard service if any and related pods
+// if any and related pods
 type TFJobDetail struct {
-	TFJob     *v1alpha1.TFJob `json:"tfJob"`
-	TbService *v1.Service     `json:"tbService"`
-	Pods      []v1.Pod        `json:"pods"`
+	TFJob *v1alpha1.TFJob `json:"tfJob"`
+	Pods  []v1.Pod        `json:"pods"`
 }
 
 // TFJobList is a list of TFJobs
@@ -140,26 +139,6 @@ func (apiHandler *APIHandler) handleGetTFJobDetail(request *restful.Request, res
 		TFJob: job,
 	}
 
-	if job.Spec.TensorBoard != nil {
-		tbSpec, err := apiHandler.cManager.ClientSet.CoreV1().Services(namespace).List(metav1.ListOptions{
-			LabelSelector: fmt.Sprintf("kubeflow.org=,app=tensorboard,runtime_id=%s", job.Spec.RuntimeId),
-		})
-		if err != nil {
-			log.Warningf("failed to list TensorBoard for TFJob %v under namespace %v, error: %v", job.Name, job.Namespace, err)
-			// TODO maybe partial result?
-			response.WriteError(http.StatusNotFound, err)
-			return
-		} else if len(tbSpec.Items) > 0 {
-			// Should never be more than 1 service that matched, handle error
-			// Handle case where no TensorBoard is found
-			tfJobDetail.TbService = &tbSpec.Items[0]
-			log.Warningf("more than one TensorBoards found for TFJob %v under namespace %v, this should be impossible",
-				job.Name, job.Namespace)
-		} else {
-			log.Warningf("Couldn't find a TensorBoard service for TFJob %v under namespace %v", job.Name, job.Namespace)
-		}
-	}
-
 	// Get associated pods
 	pods, err := apiHandler.cManager.ClientSet.CoreV1().Pods(namespace).List(metav1.ListOptions{
 		LabelSelector: fmt.Sprintf("kubeflow.org=,runtime_id=%s", job.Spec.RuntimeId),

diff --git a/dashboard/frontend/src/components/CreateJob.js b/dashboard/frontend/src/components/CreateJob.js
@@ -8,7 +8,6 @@ import RaisedButton from "material-ui/RaisedButton";
 import { withRouter } from "react-router-dom";
 
 import { createTFJobService } from "../services";
-import CreateTensorBoard from "./CreateTensorBoard";
 import RequiredTextField from "./RequiredTextField";
 import VolumeCreator from "./VolumeCreator";
 import EnvVarCreator from "./EnvVarCreator";
@@ -33,8 +32,6 @@ class CreateJob extends Component {
       psImage: "",
       psCommand: "",
       psArgs: "",
-      tbIsPresent: false,
-      tbSpec: {},
       masterVolumeSpec: {},
       workerVolumeSpec: {},
       psVolumeSpec: {},
@@ -44,7 +41,6 @@ class CreateJob extends Component {
     };
 
     this.handleInputChange = this.handleInputChange.bind(this);
-    this.setTensorboardSpec = this.setTensorboardSpec.bind(this);
     this.cancel = this.cancel.bind(this);
     this.deploy = this.deploy.bind(this);
     this.setMasterVolumesSpec = this.setMasterVolumesSpec.bind(this);
@@ -55,10 +51,6 @@ class CreateJob extends Component {
     this.setPSEnvVars = this.setPSEnvVars.bind(this);
   }
 
-  setTensorboardSpec(tbSpec) {
-    this.setState({ tbSpec });
-  }
-
   handleInputChange(event) {
     const target = event.target;
     const value = target.type === "checkbox" ? target.checked : target.value;
@@ -230,20 +222,6 @@ class CreateJob extends Component {
               <VolumeCreator setVolumesSpec={this.setPSVolumesSpec} />
             </div>
           )}
-
-          {/* TENSORBOARD */}
-          <Divider style={this.styles.divider} />
-          <h3 style={this.styles.header}>TensorBoard</h3>
-          <Toggle
-            label="Enabled"
-            defaultToggled={false}
-            name="tbIsPresent"
-            onToggle={this.handleInputChange}
-            style={this.styles.toggle}
-          />
-          {this.state.tbIsPresent && (
-            <CreateTensorBoard setTensorBoardSpec={this.setTensorboardSpec} />
-          )}
         </CardText>
         <CardActions>
           <RaisedButton label="Deploy" primary={true} onClick={this.deploy} />
@@ -302,10 +280,6 @@ class CreateJob extends Component {
       }
     };
 
-    if (this.state.tbIsPresent) {
-      spec.spec.tensorboard = this.state.tbSpec;
-    }
-
     createTFJobService(spec)
       .then(() => this.props.history.push("/"))
       .catch(console.error);

diff --git a/dashboard/frontend/src/components/CreateTensorBoard.js b/dashboard/frontend/src/components/CreateTensorBoard.js
diff --git a/dashboard/frontend/src/components/Job.js b/dashboard/frontend/src/components/Job.js
@@ -1,7 +1,6 @@
 import React, { Component } from "react";
 import JobDetail from "./JobDetail.js";
 import ReplicaSpec from "./ReplicaSpec.js";
-import TensorBoard from "./TensorBoard.js";
 import { Card, CardText } from "material-ui/Card";
 import { getTFJobService } from "../services";
 
@@ -35,9 +34,6 @@ class Job extends Component {
           <div style={this.divStyle}>
             <JobDetail tfjob={job} />
           </div>
-          <div style={this.divStyle}>
-            <TensorBoard service={this.state.tbService} />
-          </div>
           {replicaSpecs}
         </div>
       );

diff --git a/dashboard/frontend/src/components/TensorBoard.js b/dashboard/frontend/src/components/TensorBoard.js
diff --git a/developer_guide.md b/developer_guide.md
@@ -22,9 +22,6 @@ $ tree -d -I 'vendor|bin|.git'
 ├── docs
 │   └── diagrams
 ├── examples
-│   ├── charts
-│   │   └── tensorboard
-│   │       └── templates
 │   ├── crd
 │   ├── gke
 │   │   └── notebook_image