Customer  Info:


Discover Financial Services
Reported by:
Nicholas Anderson, Senior Data Engineer 
Date: June 22nd, 2020

________________________________________________________________________________


Symptoms:

able to deploy driverlessAI local with provided docker image but having issues rebuilding with our DFS rhel base image curling dai-1.8.7-1.x86_64.rpm to reproduce

Mainly: File permissions with root and the dai-env.sh. 
It takes a while to curl 1.8gb rpm every-time
on cpu. (32cpu/256gb) (so resource is not an issue)


to curl the 1.8.7 rpm, and the shell scripts to invoke it...but missing a few files in /opt/h2oai/dai/ and need to create directories in non-root paths as we can't run images on our k8 cluster as root UID.
Documentation points to loading the pre-built image. 


________________________________________________________________________________


H2O.ai


Support: Mohamed Asni, Retima Dangol

Account Executive: Josephine Wang


FIX:


-- Needed some modification on the docker image 


followed the instruction from the Mohamed Asni's repo:


https://github.com/h2oai/driverlessai-alt-containers/blob/master/centos7/Dockerfile.in


which provides examples on how to build custom Docker images on various base images.


steps on customization on the dockerfile from the repo


1. changed to their base image 

FROM --> nvidia/cuda:10.0-cudnn7-devel-centos7---> TO their base image 


2. removed the from the run block: 


line 75-78

# Remove Python 2

RUN \

    rpm -e --nodeps python


3. added proxy 


this fixed the issue and he was able to configure DAI on Discover environment, we were able to open DAI GUI, upload the data, details look on data, auto viz,  split the data, run the experiment and MLI


________________________________________________________________________________


Additional help requested by Discover:


Nick wanted addition help on:



1. LDAP authorization baked into docker image

FIX:

pam.d -- add authentication for LDAP they allow only one user per container

    LDAP--> generate pam.d file--> DAI authentic against the pam.d file


2. Snowflakes configuration --> https://support.h2o.ai/a/tickets/97349