Skip to content

Latest commit

 

History

History
executable file
·
85 lines (66 loc) · 3.13 KB

CoreOS.md

File metadata and controls

executable file
·
85 lines (66 loc) · 3.13 KB
Raw

Deploy DL workspace cluster on a cluster on an already deployed with CoreOS cluster

This document describes the procedure to deploy DL workspace on a prior deployed CoreOS cluster. This document is still evolving. Please contact the author for any question. Our latest deployment is more focused on Ubuntu, so this portion may be a bit outdated.

  1. Configuration cluster.
  • Please copy src/ClusterBootstrap/config_coreos.yaml.template to src/ClusterBootstrap/config.yaml, fill in the cluster name, number of etcd server, domain, and the machines involved in the deployment.
cluster_name : <<your_cluster_name>>
etcd_node_num : 3
network:
  domain: <<your_cluster_domain>>
  container-network-iprange: "<<your_cluster_ip_range, in 10.109.x.x/24 format>>" 

machines:
  <<infrastructure_machine_1>>:
    role: infrastructure
  <<infrastructure_machine_2>>:
    role: infrastructure
  <<worker_machine>>:
    role: worker

  • Configure and setup the databased used in the cluster.

  • Please edit cluster.yaml generated in the step above, and remove/change :

deploydockerETCD: true [change from false -> true]
platform-scripts: coreos [change from ubuntu -> coreos]
  1. [Build deployment configuration] (Build.md).
./deploy.py -y build

If the machine is imaged through the PXEServer process, you can skip this step. If the machine is imaged and provided to you, please fill in config.yaml the installation username:

admin_username: <<yourusername>>

and put password in ./deploy/sshkey/rootpasswd. The run

./deploy.py sshkey install

After this step, you should be able to access the cluster via Scripts. Please make sure that the user is in sudo and docker group.

  1. Partition hard drive, if necessary. Please refer to section Partition for details.

  2. Setup kubernetes

./deploy.py download kubectl 
./deploy.py -y deploy
./deploy.py -y updateworker
./deploy.py -y kubernetes labels

If you are running a small cluster, and need to run workload on the Kubernete master node (this choice may affect cluster stability), please use:

./deploy.py -y kubernetes uncordon

Works now will be scheduled on the master node. If you stop here, you will have a fully functional kubernete cluster. Thus, part of DL Workspace setup can be considered automatic procedure to setup a kubernete cluster. You don't need shared file system or database for kubernete cluster operation.

  1. [optional] Configure, setup GlusterFS

  2. [Optional] Configure, setup HDFS

  3. [Optional] Setup Spark

  4. Mount shared file system, please note that CoreOS can only mount NFS, you cannot mount CIFS

./deploy.py mount
  1. Start webUI service.
./deploy.py webui
./deploy.py docker build restfulapi
./deploy.py docker build webui
./deploy.py kubernetes start jobmanager
./deploy.py kubernetes start restfulapi
./deploy.py kubernetes start webportal