Skip to content

Latest commit

 

History

History
93 lines (56 loc) · 3.87 KB

distributed-deploy.md

File metadata and controls

93 lines (56 loc) · 3.87 KB

Distributed deploy

Index


Step 1. Prepare Deployment Environment

Note 1: If you wanna manage the cluster in a machine belonging to OpenPAI, please choose Option B. Otherwise, Option A is highly recommended.

Note 2: We suggest the Option A. Because manage cluster in a node of the cluster is not recommanded.


Step 2. Prepare Configuration

If this is the first time for you to deploy OpenPAI, Option A is highly recommended.

If you wanna customize the configuration, you could firstly follow the Option A to generate a rough cluster configuration. And then you can follow Option B to customize your configuration according to your requirements.

If you are very familiar with OpenPAI, you could directly write your configuration based on the Option B.


Step 3. Deploy Kubernetes

If your cluster is deployed in Azure, and there are azure rdma capable machines. Please go to this section first.

Suggest to deploy kubernetes and maintain it with kubespray. We have written a document for user to deploy a k8s cluster. And when deploy k8s through this way, please choose an VM which is not belong to the cluster.

Of course, you could deploy kubernetes through paictl. But the kubernetes doesn't have rbac and tls. And due to its risk, deploying k8s with kubespray is highly recommanded.


Step 4. Update cluster configuration into Kubernetes


Step 5. Start all OpenPAI services


Appendix. Validate deployment


Appendix. Azure RDMA