developers.redhat.com Open in urlscan Pro
2a02:26f0:6c00::210:baa2  Public Scan

Form analysis
1 forms found in the DOM

GET /search

<form action="/search" method="get">
  <label for="pfe-navigation__search-label-universal">Search</label>
  <input id="pfe-navigation__search-label-universal" name="t" type="text" placeholder="Search">
  <button>Search</button>
</form>

Text Content

Skip to main content
 * Topics
   
   
   FEATURED TOPICS
   
    * Kubernetes
      Learn how this powerful open-source tool helps you manage components
      across containers in any environment.
    * Quarkus
      Kubernetes-native Java with low memory footprint and fast boot times for
      microservices and serverless applications.
    * DevOps
      DevOps involves the combination of cultural change, process automation,
      and tools to improve your time-to-market.
    * Linux
      Develop applications on the most popular Linux for the enterprise—all
      while using the latest technologies.
   
   
   OTHER TOPICS
   
    * .NET Core
    * Apache Kafka on Kubernetes
    * API Management
    * Camel K
    * Containers
    * Data Integration
    * Data Science
    * DevOps
    * DevTools
    * Edge computing
    * Event-Driven Architecture
   
    * GitOps
    * Istio service mesh
    * Java
    * Javascript
    * Microservices
    * Node.js
    * Open Source
    * Operators
    * Python
    * Serverless
    * Spring Boot
   
    * View all topics

 * Products
   
   
   FEATURED PRODUCTS
   
    * Red Hat Enterprise Linux
      A stable, proven foundation that's versatile enough for rolling out new
      applications, virtualizing environments, and creating a secure hybrid
      cloud.
    * OpenShift
      Open, hybrid-cloud Kubernetes platform to build, run, and scale
      container-based applications -- now with developer tools, CI/CD, and
      release management.
    * Red Hat build of OpenJDK
      The Red Hat build of OpenJDK is a free and supportable open source
      implementation of the Java Platform, Standard Edition (Java SE).
   
   
   MORE PRODUCTS
   
    * 3scale API Management
    * AMQ
    * Ansible Automation Platform
    * CodeReady Containers
    * CodeReady Studio
    * CodeReady Workspaces
    * Container Development Kit
    * Fuse
    * JBoss Enterprise Application Platform
   
    * Process Automation Manager
    * Migration Toolkit for Applications
    * OpenShift API Management
    * OpenShift Application Runtimes
    * OpenShift Data Science
    * OpenShift Streams for Apache Kafka
    * Red Hat Decision Manager
    * Red Hat Developer Toolset
    * Red Hat build of Quarkus
   
    * View all products
   
   
   
    * Develop in the sandbox

 * Developer Sandbox
 * Build
 * Tools
 * Events
 * Learn
 * Partner

Search Search
All Red Hat
Back to menu

 * You are here
   
   
   
   
   RED HAT
   
   Learn about our open source products, services, and company.

 * You are here
   
   
   
   
   RED HAT CUSTOMER PORTAL
   
   Get product support and knowledge from the open source experts.

 * You are here
   
   
   
   
   RED HAT DEVELOPER
   
   Read developer tutorials and download Red Hat software for cloud application
   development.

 * You are here
   
   
   
   
   RED HAT PARTNER CONNECT
   
   Become a Red Hat partner and get support in building customer solutions.

--------------------------------------------------------------------------------


 * PRODUCTS
   
   
   * ANSIBLE.COM
     
     Learn about and try our IT automation product.


 * TRY, BUY, SELL
   
   
   * RED HAT HYBRID CLOUD
     
     Access technical how-tos, tutorials, and learning paths focused on Red
     Hat’s hybrid cloud managed services.
   
   
   * RED HAT STORE
     
     Buy select Red Hat products and services online.
   
   
   * RED HAT MARKETPLACE
     
     Try, buy, sell, and manage certified enterprise software for
     container-based environments.


 * COMMUNITY & OPEN SOURCE
   
   
   * THE ENTERPRISERS PROJECT
     
     Read analysis and advice articles written by CIOs, for CIOs.
   
   
   * OPENSOURCE.COM
     
     Read articles on a range of topics about open source.


 *  
   
   
   * RED HAT SUMMIT
     
     Register for and learn about our annual open source IT industry event.
   
   
   * RED HAT ECOSYSTEM CATALOG
     
     Find hardware, software, and cloud providers―and download container
     images―certified to perform with Red Hat technologies.

Article


HOW TO INSTALL AN OPEN SOURCE TOOL FOR CREATING MACHINE LEARNING PIPELINES

May 5, 2022
Share on twitter Share on facebook Share on linkedin Share with email

Tags:
AI/ML CI/CD
JooHo Lee

Table of contents:
 * Deploy the Open Data Hub Operator
 * Create a KfDef to deploy Pachyderm, JuypterHub, and Ceph Nano
 * Video demo
 * Troubleshooting

Pachyderm is an open source tool for creating and running machine learning
(AI/ML) pipelines. It runs on Kubernetes and provides modern developer benefits
such as versioning and autoscaling. Pachyderm also integrates with the
JupyterHub information sharing platform. On March 8, 2022, Red Hat announced
that Pachyderm has been added to the Open Data Hub (ODH), which is a blueprint
for building an AI-as-a-service platform on the Red Hat OpenShift Container
Platform. In this article, you'll learn how to install Pachyderm using Open Data
Hub.

To walk through the steps in this article, you'll need a Red Hat OpenShift
cluster with a default StorageClass. The procedure in this article has been
tested in the following environments:

 * OpenShift Dedicated 4.9 on AWS with a gp2 StorageClass
 * An OpenShift cluster using Red Hat OpenShift Local (formerly Red Hat
   CodeReady Containers) with an nfs StorageClass set up by the NFS Provisioner
   Operator.

The first option is used in this article. OpenShift Dedicated provides a default
gp2 StorageClass, but it is not cost-free. As an alternative, with the second
option you can set up a cost-free environment as follows:

 1. Use OpenShift Local to install an OpenShift All-in-One cluster on your
    laptop.
 2. Add an nfs StorageClass using the NFS Provisioner Operator, available from
    OperatorHub or Github.

Once you've followed those steps, you'll have essentially the same environment
as OpenShift Dedicated.

This article also contains an embedded video illustrating the steps.

If you want to experiment with the Red Hat OpenShift Local test environment,
please refer to the following articles:

 * Configure CodeReady Containers for AI/ML development
 * Create and manage local persistent volumes with CodeReady Containers

Deploy the Open Data Hub Operator


DEPLOY THE OPEN DATA HUB OPERATOR

Installing an Operator is the easiest step in this procedure. Go to the
OperatorHub menu option in the OpenShift console, search for the Open Data Hub
Operator, and click its link (Figure 1).

Figure 1. From the OperatorHub menu option, search for the Open Data Hub
Operator and click its link.
Figure 1: Find the Open Data Hub Operator in the Openshift console.

You'll be taken to the Operator page for the Open Data Hub Operator (Figure 2).
Click Install.

Figure 2. From the Open Data Hub Operator page, click Install.
Figure 2: Begin the Operator installation process by clicking here.

Next, you'll see the Install Operator page (Figure 3). Keep all the defaults and
click Install again.

Figure 3. From the Install Operator page, click Install.
Figure 3: Click Install on the Install Operator page.

When installation is complete, you'll see a message saying "saying "Installed
operator — ready for use," as in Figure 4.

Figure 4. When you finish installation, a page comes up saying "Installed
operator — ready for use."
Figure 4: The Operator is now installed.

After you install Open Data Hub Operator, you need to create a new project that
we'll call opendatahub, where all required components—Jupyterhub, Ceph Nano, and
Pachyderm—will be deployed

$ oc new-project opendatahub

Copy snippet
Create a KfDef to deploy Pachyderm, JuypterHub, and Ceph Nano


CREATE A KFDEF TO DEPLOY PACHYDERM, JUYPTERHUB, AND CEPH NANO

Pachyderm supports any storage option compatible with AWS S3 object storage.
Open Data Hub provides two of these storage options:

 * Full automation: Deploy Ceph Nano on Open Data Hub, which creates a secret
   for Pachyderm.
 * Partial automation: Manually create a secret for the credentials to access S3
   or another S3-compatible object storage, such as MinIO.


FULL AUTOMATION USING CEPH NANO

Open Data Hub provides a full automation YAML configuration using a Kubernetes
Job named pachyderm-deployer. Here's an excerpt of the configuration:

# Ceph Nano 
- kustomizeConfig:
    repoRef:
      name: manifests
      path: ceph/object-storage/scc
  name: ceph-nano-scc
- kustomizeConfig:
    repoRef:
      name: manifests
      path: ceph/object-storage/nano
  name: ceph-nano

# Pachyderm operator
- kustomizeConfig:
    parameters:
      - name: namespace
        value: openshift-operators
    repoRef:
      name: manifests
      path: odhpachyderm/operator
  name: odhpachyderm-operator

# Pachyderm deployer
- kustomizeConfig:
    repoRef:
      name: manifests
      path: odhpachyderm/deployer
  name: odhpachyderm-deployer

Copy snippet

The configuration contains a script that makes sure Ceph Nano is in a ready
state, and then creates an S3 bucket in Ceph Nano. After that, the script
creates a secret for the S3 bucket credentials, which Pachyderm will use to gain
access to the S3 bucket.

To use full automation on Kubernetes, you need a KfDef custom resource (CR). A
manifest for this KfDef can be found in my GitHub repository. Create the KfDef
on OpenShift through the following command:

$ oc create -f https://bit.ly/3wHwt59

Copy snippet


PARTIAL AUTOMATION USING S3 OR OTHER COMPATIBLE STORAGE (MINIO)

If you want to go the partial automation route, the only difference from using
Ceph Nano is that you need to create a secret before creating the KfDef, then
pass that information to pachyderm-deployer in the KfDef. The relevant line can
be found in context in the YAML file.

An oc command that creates a secret for AWS S3 looks like this:

$ oc create secret generic pachyderm-aws-secret \
--from-literal=access-id=XXX  \
--from-literal=access-secret=XXX \
--from-literal=region=us-east-2 \
--from-literal=bucket=pachyderm 

Copy snippet

An oc command that creates a secret for MinIO looks like this:

$ oc create secret generic pachyderm-minio-secret \
--from-literal=access-id=XXX  \
--from-literal=access-secret=XXX \
--from-literal=custom-endpoint=${minio_ip}
--from-literal=region=us-east-2 \
--from-literal=bucket=pachyderm 

Copy snippet

The following excerpt from a KfDef manifest shows how to use the secret with S3.
The example uses pachyderm-aws-secret for the secret:

# Pachyderm Operator
- kustomizeConfig:
    parameters:
      - name: namespace
        value: openshift-operators
    repoRef:
      name: manifests
      path: odhpachyderm/operator
  name: odhpachyderm-operator

# Pachyderm Deployer
- kustomizeConfig:
    parameters:
      - name: storage_secret             #<=== Must set this
        value: pachyderm-aws-secret      #<=== Use your Secret Name
    repoRef:
      name: manifests
      path: odhpachyderm/deployer
  name: odhpachyderm-deployer

Copy snippet

Once you've created the secret, you can create the KfDef through the following
command:

$ oc create -f https://bit.ly/3NkV31I

Copy snippet

After you create the KfDef, OpenShift creates several pods in the opendatahub
project. Four pods are created for Pachyderm:

$ oc get pod 
etcd-0                          1/1     Running   0          12m
postgres-0                      1/1     Running   0          12m
pachd-874f5958c-7w98p           1/1     Running   0          11m
pg-bouncer-7587d49769-gwn8f     1/1     Running   0          11m

Copy snippet

Even more pods might be devoted to Pachyderm if you are using Red Hat OpenShift
Local. If resources on your cluster are tight, it could take some time to create
the pods.

Now you can try Pachyderm on your cluster.

Video demo


VIDEO DEMO

The following video illustrates the steps outlined so far.


Troubleshooting


TROUBLESHOOTING

If you are running this example on your laptop, you might see some errors with
the JupyterHub pods, jupyterhub and jupyterhub-db, due to a lack of resources
(Figure 5).

Figure 5. Sometimes, jupyterhub and jupyterhub-db show errors at startup.
Figure 5: Sometimes, jupyterhub and jupyterhub-db show errors at startup.

traefix-proxy pods might show some errors, but you can ignore them. When
jupyterhub and jupyterhub-db are recovered, traefix-proxy will be automatically
healed.

If you see these errors, start a rollout of the DeploymentConfigs for jupyterhub
and jupyterhub-db as shown in Figures 6 and 7.

Figure 6. Pull up the DeploymentConfigs page to get access to pages for
jupyterhub and jupyterhub-dc.
Figure 6: Pull up the DeploymentConfigs page to get access to pages for
jupyterhub and jupyterhub-dc.
Figure 7. On the jupyterhub page, choose "Start rollout."
Figure 7: On the jupyterhub page, choose Start rollout.

Start a rollout for jupyterhub-db in the same way.

If these steps don't solve the problem, roll out jupyterhub-db first, wait until
it is ready, then roll out jupyterhub. Then enjoy experimenting with what
Pachyderm has to offer!


RECENT ARTICLES


 * ALL ABOUT LOCAL AND SELF-MANAGED KAFKA DISTRIBUTIONS


 * HOW TO USE OPERATORS WITH AWS CONTROLLERS FOR KUBERNETES


 * DEVELOPER TOOLS REBRAND, SAY FAREWELL TO CODEREADY NAME


 * HOW TO ORGANIZE JFR DATA WITH RECORDING LABELS IN CRYOSTAT 2.1


 * RHEL 8.6: WHAT'S NEW AND HOW TO UPGRADE


RELATED CONTENT


CONFIGURE CODEREADY CONTAINERS FOR AI/ML DEVELOPMENT


WHAT’S UP NEXT?

 

Open Source Data Pipelines for Intelligent Applications provides data engineers
and scientists insight into how Kubernetes provides a platform for building data
platforms that increase an organization’s data agility. 

Download the free e-book


COMMENTS


Please enable JavaScript to view the comments powered by Disqus.


 * FEATURED TOPICS
   
   * Istio
   * Quarkus
   * CI/CD
   * Serverless
   * Enterprise Java
   * Linux
   * Microservices
   * DevOps


 * BUILD
   
   * Getting Started Center
   * Developer Tools
   * Interactive Tutorials
   * Container Catalog
   * Operators Marketplace
   * Certify Applications
   * Red Hat on Github


 * QUICKLINKS
   
   * What's new
   * DevNation events
   * Upcoming Events
   * Books
   * Cheat Sheets
   * Videos
   * Products


 * COMMUNICATE
   
   * Site Status Dashboard
   * Report a website issue
   * Report a security problem
   * Helping during COVID-19
   * About us
   * Contact Sales

Red Hat Developer

Build here. Go anywhere.

We serve the builders. The problem solvers who create careers with code.

Join us if you’re a developer, software engineer, web designer, front-end
designer, UX designer, computer scientist, architect, tester, product manager,
project manager or team lead.

 * 
 * 
 * 
 * 

Sign me up
 * ©2022 Red Hat, Inc.
 * Cookie-Präferenzen
 * Privacy Statement
 * Terms of Use
 * All policies and guidelines


✓
Thanks for sharing!
AddToAny
More…