Skip to content

Continue a Project

Jump to bottom
Tom Farrand edited this page Mar 13, 2020 · 1 revision

Create a new WMLCE environment, attached to an existing block storage, in an existing VPC

This guide will walk you through how to add this git repo to IBM Cloud Schematics. Once added, the schematic can be deployed to install a dedicated instance of Watson Machine Learning Community Edition attached to an existing block storage volume inside of an existing Virtual Private Cloud.

All of the steps are identical to those described here apart from step 4, therefore this tutorial should be followed in advance of this one. Alternatively, if you have a VPC setup and are familiar with IBM Cloud Schematics already you should be able to follow regardless.

This wiki assumes you have successfully navigated to the IBM Cloud Schematics page, and are creating a new workspace. You should have your IBM Cloud API key to hand- see here for more info.

The majority of the input variables which need to be entered into the new workspace are the same as those used to create the VPC initially, however these is a subset of new variables which are highlighted in the first table below.

Input Variable Description
subnet_id ID of the existing subnet to attach onto. All of your subnets will be listed here: https://cloud.ibm.com/vpc-ext/network/subnets
block_storage_volume_id The block storage volume ID within your VPC where user data is persisted. All of your block storage devices will be listed here: https://cloud.ibm.com/vpc-ext/storage/storageVolumes

Variables which are the same as before:

Input Variable Value Sensitive?
ibmcloud_api_key Yes
vpc_basename wmlce-vpc (Tailor to something more memorable)

Other important input variables are displayed below. These can be left with their default values, but to tailor the infrastructure which is provisioned to your needs they should be customised.

Input Variable Description
vm_name Name of the VSI upon which WMLCE will be installed.
vm_profile The compute power assigned to the VSI hosting WMLCE. GPU compute profiles can be viewed here: https://cloud.ibm.com/docs/vpc?topic=vpc-profiles#gpu

Next create your new plan, by selecting Create on the right hand pane. Finally, you will want to apply your plan. Click Apply Plan to kick this process off.

Apply plan button

Alternately you can do a "dry run" of the plan by choosing Generate Plan. This will show you which infrastructure is going to be provisioned by your plan.

Now that your plan is being applied you can view the logs live. The whole process will take approximately 15 minutes.

Once the plan has been applied the logs will display three key pieces of information:

  • The internet facing URL of your Jupyterlab workbench e.g. https://169.48.155.123:443/lab
  • The randomly generated access token e.g. vMCH82fVYNzTLLTW
  • The private RSA key for the VSI. This will allow you to SSH into the box either to debug or administer the system. If you need to ssh into the instance you may paste the key into a file called "private.key" set its permissions to "600" and then SSH from your laptop to the IP address. e.g.
$ vim private.key
$ chmod 600 private.key
$  ssh -i private.key root@169.48.154.123
Welcome to Ubuntu 18.04.3 LTS (GNU/Linux 4.15.0-66-generic ppc64le)

 * Documentation:  https://help.ubuntu.com
 * Management:     https://landscape.canonical.com
 * Support:        https://ubuntu.com/advantage

  System information as of Wed Jan  8 21:33:22 CST 2020

  System load:  0.16              Processes:              590
  Usage of /:   6.3% of 97.92GB   Users logged in:        0
  Memory usage: 3%                IP address for enp0s1:  10.240.0.4
  Swap usage:   0%                IP address for docker0: 172.17.0.1

 * Overheard at KubeCon: "microk8s.status just blew my mind".

     https://microk8s.io/docs/commands#microk8s.status

66 packages can be updated.
28 updates are security updates.

Your Hardware Enablement Stack (HWE) is supported until April 2023.

Last login: Wed Jan  8 11:28:29 2020 from 136.49.83.203
root@vision-prod-vm1:~#

Destroying Resources

To destroy all of the resources associated with this plan, you can select Actions -> Delete and then remove all associated resources with the workspace and/or delete the workspace itself.

Debugging and Troubleshooting

Debugging should be done readily through the logging interface. If the provisioning fails, most of the time it is a temporary error, and so the best thing to do is to simply delete the associated resources, and re-apply the plan.

If that does not work be sure to go back through and check you have correctly filled in the variables. Ensure that none of your variable names include an underscore, as I have had issues with this in the past.

Finally, if you are still stuck please open an issue and I will get back to you ASAP!

All of these actions are also accessible through REST APIs, or the ibmcloud CLI. This can help to automate the provisioning of infrastructure inside of production workflows.

Clone this wiki locally