Skip to content

Latest commit

 

History

History
101 lines (61 loc) · 5.13 KB

deep-water-ami.md

File metadata and controls

101 lines (61 loc) · 5.13 KB
Raw

Deep Water AMI

An Amazon Machine Image (AMI) provides a convenient way for H2O to distribute Deep Water. Deep Water AMIs are produced periodcally, often based on a known stable AMI previously produced. When an AMI is made public, H2O.ai does not support the AMI for enterprise or production use cases. AMIs are merely provided for user experiments and training.

AMIs are specific to Amazon Elastic Compute Cloud (EC2) regions and, as such, AMI IDs are unique to regions. Be sure that the AMI ID you use is for the correct region.

By default, Amazon AMIs update packages. This can cause problems with GPU drivers, which are tied to the OS kernel. Updates to the kernel can cause a mismatch between the installed GPU drivers and break the Deep Water stack. Because of this, we recommend that you:

  • Prevent automatic updates using: sudo apt-get remove unattended-upgrades

    or

  • Re-install the drivers to match the new kernel.

Deep Water Workshop on EC2 Amazon AMI

A Deep Water workshop was presented at H2O Open Tour Dallas. The hands-on workshop is available in a public EC2 Amazon AMI. This document describes how to load and run this workshop. Note that this requires an account on Amazon AWS.

  1. Log in to your your AWS account at https://aws.amazon.com.

  2. In the upper right corner of the Amazon Web Services page, change the location in the location drop-down to US East (N Virginia).

    Change Location

  3. Select the EC2 option under the Compute section to launch the EC2 Dashboard.

    Select EC2

  4. Select Images > AMIs on left navigation.

    Images > AMIs

  5. On the Launch screen, change the dropdown at the top to Public images , then search for the Deep Water AMI using the ID: ami-10bd9607. Click Enter to begin the search.

    Locate the AMI

  6. After the AMI is located, click the Launch button.

  7. At this point, you will be directed to choose your GPU instance type. Select an instance, for example g2.2xlarge, then click Next: Configure Instance Details.

    Choose instance type

  8. Accept the default configuration for this instance, then click Next: Add Storage.

    Configure instance details

  9. Specify a value greater than or equal to 50 GB for the Size value (storage size), then click Next: Tag Instance.

    Add storage

  10. Enter a unique name tag to identify your instance, then click Next: Configure Security Group.

    Tag instance

  11. Update the security group, and add rules as indicated in the following table (refer also to the image below the table):

    | Type            | Protocol  | Port Range    | Source             |
| ----------------|-----------|---------------|--------------------|
| SSH             | TCP       | 22            | Anywhere 0.0.0.0/0 |      
| HTTP            | TCP       | 80            | Anywhere 0.0.0.0/0 |  
| HTTPS           | TCP       | 443           | Anywhere 0.0.0.0/0 | 
| Custom TCP Rule | TCP       | 8080          | Anywhere 0.0.0.0/0 |
| Custom TCP Rule | TCP       | 54321-54330   | Anywhere 0.0.0.0/0 | 
| Custom TCP Rule | TCP       | 55001         | Anywhere 0.0.0.0/0 | 
| Custom TCP Rule | TCP       | 55011         | Anywhere 0.0.0.0/0 |
| Custom TCP Rule | TCP       | 55021         | Anywhere 0.0.0.0/0 |

    These rules are necessary to open the Flow UI, Prediction Services, Jupyter Notebook server and log in to the instance via command line. Click Review and Launch.

    Configure security group

  12. Review the configuration, and then click Launch.

    Review instance launch

  13. A popup will appear prompting you to select a key pair. This will be used to log in to the instance via command line. You can select your existing key pair or create a new one. Be sure to accept the acknowledgement, then click Launch Instances to start the new instance.

    Enter or select key/pair

After the instance starts, you can view/start/stop/terminate the instance from the EC2 Dashboard by clicking on Running Instances.

  • To open a Jupyter Notebook server, enter <Public_IP_Address>:80 in the address bar of your browser. A message box will appear, prompting you to provide authentication. Enter deepwater for the username, and enter the AWS Instance ID as the password.

  • To open the Flow UI, enter <Public_IP_Address>:54321 in the address bar of your browser.

  • You can log in to this instance using ssh with Terminal (Mac/Linux) or Putty (Window). For example:

      ssh -i <Private_Key_File> ubuntu@<Public_IP_Address>``

Note that the public IP address will change on reboot. Also, the key pair file should have restricted permissions (chmod 400 <Private_Key_File>).