volcano.sh/volcano@v1.9.0/test/e2e/jobseq/tensorflow.go (about)

     1  /*
     2  Copyright 2021 The Volcano Authors.
     3  
     4  Licensed under the Apache License, Version 2.0 (the "License");
     5  you may not use this file except in compliance with the License.
     6  You may obtain a copy of the License at
     7  
     8      http://www.apache.org/licenses/LICENSE-2.0
     9  
    10  Unless required by applicable law or agreed to in writing, software
    11  distributed under the License is distributed on an "AS IS" BASIS,
    12  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
    13  See the License for the specific language governing permissions and
    14  limitations under the License.
    15  */
    16  
    17  package jobseq
    18  
    19  import (
    20  	"context"
    21  
    22  	. "github.com/onsi/ginkgo/v2"
    23  	. "github.com/onsi/gomega"
    24  
    25  	v1 "k8s.io/api/core/v1"
    26  	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
    27  
    28  	vcbatch "volcano.sh/apis/pkg/apis/batch/v1alpha1"
    29  	vcbus "volcano.sh/apis/pkg/apis/bus/v1alpha1"
    30  
    31  	e2eutil "volcano.sh/volcano/test/e2e/util"
    32  )
    33  
    34  var _ = Describe("TensorFlow E2E Test", func() {
    35  	It("Will Start in pending state and goes through other phases to get complete phase", func() {
    36  		ctx := e2eutil.InitTestContext(e2eutil.Options{})
    37  		defer e2eutil.CleanupTestContext(ctx)
    38  
    39  		jobName := "tensorflow-dist-mnist"
    40  
    41  		job := &vcbatch.Job{
    42  			ObjectMeta: metav1.ObjectMeta{
    43  				Name: jobName,
    44  			},
    45  			Spec: vcbatch.JobSpec{
    46  				MinAvailable:  int32(3),
    47  				SchedulerName: e2eutil.SchedulerName,
    48  				Plugins: map[string][]string{
    49  					"svc": {},
    50  					"env": {},
    51  				},
    52  				Policies: []vcbatch.LifecyclePolicy{
    53  					{
    54  						Event:  vcbus.PodEvictedEvent,
    55  						Action: vcbus.RestartJobAction,
    56  					},
    57  				},
    58  				Tasks: []vcbatch.TaskSpec{
    59  					{
    60  						Replicas: int32(1),
    61  						Name:     "ps",
    62  						Template: v1.PodTemplateSpec{
    63  							Spec: v1.PodSpec{
    64  								RestartPolicy: v1.RestartPolicyNever,
    65  								Containers: []v1.Container{
    66  									{
    67  										Command: []string{
    68  											"sh",
    69  											"-c",
    70  											"PS_HOST=`cat /etc/volcano/ps.host | sed 's/$/&:2222/g' | sed 's/^/\"/;s/$/\"/' | tr \"\n\" \",\"`; WORKER_HOST=`cat /etc/volcano/worker.host | sed 's/$/&:2222/g' | sed 's/^/\"/;s/$/\"/' | tr \"\n\" \",\"`; export TF_CONFIG={\\\"cluster\\\":{\\\"ps\\\":[${PS_HOST}],\\\"worker\\\":[${WORKER_HOST}]},\\\"task\\\":{\\\"type\\\":\\\"ps\\\",\\\"index\\\":${VK_TASK_INDEX}},\\\"environment\\\":\\\"cloud\\\"}; echo ${TF_CONFIG}; python /var/tf_dist_mnist/dist_mnist.py --train_steps 1000",
    71  										},
    72  										Image: e2eutil.DefaultTFImage,
    73  										Name:  "tensorflow",
    74  										Ports: []v1.ContainerPort{
    75  											{
    76  												Name:          "tfjob-port",
    77  												ContainerPort: int32(2222),
    78  											},
    79  										},
    80  									},
    81  								},
    82  							},
    83  						},
    84  					},
    85  					{
    86  						Replicas: int32(2),
    87  						Name:     "worker",
    88  						Policies: []vcbatch.LifecyclePolicy{
    89  							{
    90  								Event:  vcbus.TaskCompletedEvent,
    91  								Action: vcbus.CompleteJobAction,
    92  							},
    93  						},
    94  						Template: v1.PodTemplateSpec{
    95  							Spec: v1.PodSpec{
    96  								RestartPolicy: v1.RestartPolicyNever,
    97  								Containers: []v1.Container{
    98  									{
    99  										Command: []string{
   100  											"sh",
   101  											"-c",
   102  											"PS_HOST=`cat /etc/volcano/ps.host | sed 's/$/&:2222/g' | sed 's/^/\"/;s/$/\"/' | tr \"\n\" \",\"`; WORKER_HOST=`cat /etc/volcano/worker.host | sed 's/$/&:2222/g' | sed 's/^/\"/;s/$/\"/' | tr \"\n\" \",\"`; export TF_CONFIG={\\\"cluster\\\":{\\\"ps\\\":[${PS_HOST}],\\\"worker\\\":[${WORKER_HOST}]},\\\"task\\\":{\\\"type\\\":\\\"worker\\\",\\\"index\\\":${VK_TASK_INDEX}},\\\"environment\\\":\\\"cloud\\\"}; echo ${TF_CONFIG}; python /var/tf_dist_mnist/dist_mnist.py --train_steps 1000",
   103  										},
   104  										Image: e2eutil.DefaultTFImage,
   105  										Name:  "tensorflow",
   106  										Ports: []v1.ContainerPort{
   107  											{
   108  												Name:          "tfjob-port",
   109  												ContainerPort: int32(2222),
   110  											},
   111  										},
   112  									},
   113  								},
   114  							},
   115  						},
   116  					},
   117  				},
   118  			},
   119  		}
   120  
   121  		created, err := ctx.Vcclient.BatchV1alpha1().Jobs(ctx.Namespace).Create(context.TODO(), job, metav1.CreateOptions{})
   122  		Expect(err).NotTo(HaveOccurred())
   123  
   124  		err = e2eutil.WaitJobStates(ctx, created, []vcbatch.JobPhase{vcbatch.Pending, vcbatch.Running, vcbatch.Completed}, e2eutil.FiveMinute)
   125  		Expect(err).NotTo(HaveOccurred())
   126  	})
   127  
   128  })