安装

上手 Volcano 最容易的方式是从 github 下载release ，然后按照以下步骤操作：

准备

  • 一个 Kubernetes 集群，集群版本不低于 V1.13
  • [可选项] 在您的集群中下载 Helm，您可以根据以下指南安装 Helm：安装 Helm。(仅当您使用 helm 模式进行安装时需要)
  • 下载最新版本 volcano下载链接
  • 解压 release 文件
  1.   #tar -xvf volcano-{Version}-linux-gnu.tar.gz
  3.   #cd volcano-{Version}-linux-gnu
  5.   #ll
  6.   total 60
  7.   drwxr-xr-x  4 root1 root1  4096 Jul 23 11:38 ./
  8.   drwxr-xr-x 11 root1 root1 12288 Jul 23 11:37 ../
  9.   drwxr-xr-x  3 root1 root1  4096 Jul 16 16:15 bin/
  10.   -rw-r--r--  1 root1 root1   153 Jul 16 16:15 default-queue.yaml
  11.   drwxr-xr-x  3 root1 root1  4096 Jul 16 16:15 helm/
  12.   -rw-r--r--  1 root1 root1  4169 Jul 16 16:15 README.md
  13.   -rw-r--r--  1 root1 root1 23453 Jul 16 16:15 volcano-{Version}.yaml

安装方式

通过 Deployment Yaml 文件方式安装

使用 release 内的文件volcano-{Version}.yaml创建 deployment。

  1. # kubectl apply -f volcano-{Version}.yaml
  2. namespace/volcano-system created
  3. configmap/volcano-scheduler-configmap created
  4. serviceaccount/volcano-scheduler created
  5. clusterrole.rbac.authorization.k8s.io/volcano-scheduler created
  6. clusterrolebinding.rbac.authorization.k8s.io/volcano-scheduler-role created
  7. deployment.apps/volcano-scheduler created
  8. serviceaccount/volcano-admission created
  9. clusterrole.rbac.authorization.k8s.io/volcano-admission created
  10. clusterrolebinding.rbac.authorization.k8s.io/volcano-admission-role created
  11. deployment.apps/volcano-admission created
  12. service/volcano-admission-service created
  13. job.batch/volcano-admission-init created
  14. serviceaccount/volcano-controllers created
  15. clusterrole.rbac.authorization.k8s.io/volcano-controllers created
  16. clusterrolebinding.rbac.authorization.k8s.io/volcano-controllers-role created
  17. deployment.apps/volcano-controllers created
  18. customresourcedefinition.apiextensions.k8s.io/jobs.batch.volcano.sh created
  19. customresourcedefinition.apiextensions.k8s.io/commands.bus.volcano.sh created
  20. customresourcedefinition.apiextensions.k8s.io/podgroups.scheduling.incubator.k8s.io created
  21. customresourcedefinition.apiextensions.k8s.io/queues.scheduling.incubator.k8s.io created
  22. customresourcedefinition.apiextensions.k8s.io/podgroups.scheduling.sigs.dev created
  23. customresourcedefinition.apiextensions.k8s.io/queues.scheduling.sigs.dev created

验证 Volcano 各组件的运行状态

  1. # kubectl get all -n volcano-system
  2. NAME                                       READY   STATUS      RESTARTS   AGE
  3. pod/volcano-admission-5bd5756f79-p89tx     1/1     Running     0          6m10s
  4. pod/volcano-admission-init-d4dns           0/1     Completed   0          6m10s
  5. pod/volcano-controllers-687948d9c8-bd28m   1/1     Running     0          6m10s
  6. pod/volcano-scheduler-94998fc64-9df5g      1/1     Running     0          6m10s
  9. NAME                                TYPE        CLUSTER-IP     EXTERNAL-IP   PORT(S)   AGE
  10. service/volcano-admission-service   ClusterIP   10.96.140.22   <none>        443/TCP   6m10s
  13. NAME                                  READY   UP-TO-DATE   AVAILABLE   AGE
  14. deployment.apps/volcano-admission     1/1     1            1           6m10s
  15. deployment.apps/volcano-controllers   1/1     1            1           6m10s
  16. deployment.apps/volcano-scheduler     1/1     1            1           6m10s
  18. NAME                                             DESIRED   CURRENT   READY   AGE
  19. replicaset.apps/volcano-admission-5bd5756f79     1         1         1       6m10s
  20. replicaset.apps/volcano-controllers-687948d9c8   1         1         1       6m10s
  21. replicaset.apps/volcano-scheduler-94998fc64      1         1         1       6m10s
  25. NAME                               COMPLETIONS   DURATION   AGE
  26. job.batch/volcano-admission-init   1/1           28s        6m10s

一切配置就绪，您可以开始使用 Volcano 部署 AI/ML 和大数据负载了。

使用 Helm 安装

如果您想使用 Helm 部署 Volcano，请先确认已经在您的集群中安装了Helm。

步骤 1：

创建一个新的命名空间。

  1. # kubectl create namespace volcano-system
  2. namespace/volcano-system created
步骤 2：

使用 Helm 进行安装。

  1. # helm install helm/chart/volcano --namespace volcano-system --name volcano
  2. NAME:   volcano
  3. LAST DEPLOYED: Tue Jul 23 20:07:29 2019
  4. NAMESPACE: volcano-system
  5. STATUS: DEPLOYED
  7. RESOURCES:
  8. ==> v1/ClusterRole
  9. NAME                 AGE
  10. volcano-admission    1s
  11. volcano-controllers  1s
  12. volcano-scheduler    1s
  14. ==> v1/ClusterRoleBinding
  15. NAME                      AGE
  16. volcano-admission-role    1s
  17. volcano-controllers-role  1s
  18. volcano-scheduler-role    1s
  20. ==> v1/ConfigMap
  21. NAME                         DATA  AGE
  22. volcano-scheduler-configmap  2     1s
  24. ==> v1/Deployment
  25. NAME                 READY  UP-TO-DATE  AVAILABLE  AGE
  26. volcano-admission    0/1    1           0          1s
  27. volcano-controllers  0/1    1           0          1s
  28. volcano-scheduler    0/1    1           0          1s
  30. ==> v1/Job
  31. NAME                    COMPLETIONS  DURATION  AGE
  32. volcano-admission-init  0/1          1s        1s
  34. ==> v1/Pod(related)
  35. NAME                                  READY  STATUS             RESTARTS  AGE
  36. volcano-admission-b45b7b76-84jmw      0/1    ContainerCreating  0         1s
  37. volcano-admission-init-fw47j          0/1    ContainerCreating  0         1s
  38. volcano-controllers-5f66f8d76c-27584  0/1    ContainerCreating  0         1s
  39. volcano-scheduler-bb4467966-z642p     0/1    Pending            0         1s
  41. ==> v1/Service
  42. NAME                       TYPE       CLUSTER-IP      EXTERNAL-IP  PORT(S)  AGE
  43. volcano-admission-service  ClusterIP  10.107.128.208  <none>       443/TCP  1s
  45. ==> v1/ServiceAccount
  46. NAME                 SECRETS  AGE
  47. volcano-admission    1        1s
  48. volcano-controllers  1        1s
  49. volcano-scheduler    1        1s
  51. ==> v1beta1/CustomResourceDefinition
  52. NAME                           AGE
  53. podgroups.scheduling.sigs.dev  1s
  54. queues.scheduling.sigs.dev     1s
  57. NOTES:
  58. Thank you for installing volcano.
  60. Your release is named volcano.
  62. For more information on volcano, visit:
  63. https://volcano.sh/
步骤 3：

验证 Volcano 各组件的运行状态。

  1. # kubectl get all -n volcano-system
  2. NAME                                       READY   STATUS              RESTARTS   AGE
  3. pod/volcano-admission-b45b7b76-84jmw       0/1     ContainerCreating   0          4m42s
  4. pod/volcano-admission-init-fw47j           0/1     ImagePullBackOff    0          4m42s
  5. pod/volcano-controllers-5f66f8d76c-27584   0/1     ImagePullBackOff    0          4m42s
  6. pod/volcano-scheduler-bb4467966-z642p      0/1     ImagePullBackOff    0          4m42s
  9. NAME                                TYPE        CLUSTER-IP       EXTERNAL-IP   PORT(S)   AGE
  10. service/volcano-admission-service   ClusterIP   10.107.128.208   <none>        443/TCP   4m42s
  13. NAME                                  READY   UP-TO-DATE   AVAILABLE   AGE
  14. deployment.apps/volcano-admission     0/1     1            0           4m42s
  15. deployment.apps/volcano-controllers   0/1     1            0           4m42s
  16. deployment.apps/volcano-scheduler     0/1     1            0           4m42s
  18. NAME                                             DESIRED   CURRENT   READY   AGE
  19. replicaset.apps/volcano-admission-b45b7b76       1         1         0       4m42s
  20. replicaset.apps/volcano-controllers-5f66f8d76c   1         1         0       4m42s
  21. replicaset.apps/volcano-scheduler-bb4467966      1         1         0       4m42s
  25. NAME                               COMPLETIONS   DURATION   AGE
  26. job.batch/volcano-admission-init   0/1           4m42s      4m42s

现在您已经完成了 Volcano 的全部安装，您可以运行如下的例子测试安装的正确性：样例