Pachyderm Docs: Beginner Tutorial

Welcome to the beginner tutorial for Pachyderm! This tutorial should take about 15 minutes to complete and introduce you to Pachyderm’s fundamental concepts.

Prerequisites #

This guide assumes that you have Pachyderm running.

For an easy and quick start, install Pachyderm on your local machine as described in our Local Installation page and start experimenting.
Or check out our Quick Install page to deploy on your favorite cloud.

💡

If you are new to Pachyderm, try Pachyderm Shell. This handy tool suggests pachctl commands as you type and helps you learn Pachyderm faster.

For this tutorial, you will use pachctl to interact with your Pachyderm cluster from your terminal window and Console (Pachyderm Web UI) to interactively visualize and explore your pipelines, your data, debug jobs, read logs, etc…

If you deployed Pachyderm locally using the default local installation instructions, you have also deployed Pachyderm Web UI. Point your browser to localhost:4000 to connect to Console. You should land on this page:

Console Landing Page

Click on your View Project (We are working on allowing you to organize your pipelines by Projects) to get started. You are all set to have a follow-along visual experience of the coming steps.

Image processing with OpenCV #

This tutorial walks you through the deployment of a Pachyderm pipeline that performs edge detection on a few images. Thanks to Pachyderm’s built-in processing primitives, we can keep our code simple but still run the pipeline in a distributed, streaming fashion. Moreover, as new data is added, the pipeline automatically processes it and outputs the results.

If you hit any errors not covered in this guide, get help in our public community Slack, submit an issue on GitHub, or email us at support@pachyderm.io. We are here to help!

Create a Repo #

A repo is the highest level data primitive in Pachyderm. Like many things in Pachyderm, it shares its name with a primitive in Git and is designed to behave analogously. Generally, repos should be dedicated to a single source of data such as log messages from a particular service, a users table, or training data for an ML model. Repos are easy to create and do not take much space when empty so do not worry about making tons of them.

📖

More about the concepts of Repository and Branch in Pachyderm.

For this demo, we create a repo called images to hold the data we want to process:

pachctl create repo images

Verify that the repository was created:

pachctl list repo

System response:

NAME   CREATED       SIZE (MASTER) ACCESS LEVEL
images 4 seconds ago ≤ 0B          [repoOwner]

This output shows that the repo has been successfully created. Because we have not added anything to it yet, the size of the repository HEAD commit on the master branch is 0B.

Check your Console and notice the creation of your repository.

ℹ️

Note the “plus” icon in your images repository. It indicates that this repository is an input repository instead of an output repository where the product of your pipeline transformation will be committed. Users can write data to input repositories only.

Console images repo

Adding Data to Pachyderm #

Now that we have created a repo, it is time to add some data. In Pachyderm, you write data to an explicit commit. Commits are immutable snapshots of your data which give Pachyderm its version control properties. You can add, remove, or update files in a given commit.

📖

More about the concept of Commit in Pachyderm.

Let’s start by adding a file, in this case an image, to a new commit. We have provided some sample images for you that we host on Imgur.

Use the pachctl put file command along with the -f flag. The -f flag can take either a local file, a URL, or a object storage bucket which it scrapes automatically. In this case, we simply pass the URL.

Unlike Git, commits in Pachyderm must be explicitly started and finished as they can contain huge amounts of data and we do not want that much dirty data hanging around in an unpersisted state. pachctl put file automatically starts and finishes a commit for you so you can add files more easily. If you want to add many files over a period of time, you can do pachctl start commit and pachctl finish commit yourself.

To commit the file liberty.png to the master branch of the images repo, run :

pachctl put file images@master:liberty.png -f http://imgur.com/46Q8nDz.png

To make sure that the data we just added is in Pachyderm.

Use the pachctl list repo command to check that data has been added:

pachctl list repo

System response:

NAME   CREATED       SIZE (MASTER) ACCESS LEVEL
images 2 minutes ago ≤ 57.27KiB    [repoOwner]

View the commit that was just created:

pachctl list commit images

System response:

REPO   BRANCH COMMIT                           FINISHED       SIZE     ORIGIN DESCRIPTION
images master 89a5ab3a23c949949f763943dd7a8aac 55 seconds ago 57.27KiB USER

View the file in that commit:

pachctl list file images@master

System response:

NAME         TYPE SIZE
/liberty.png file 57.27KiB

In your Console, click on the images repo to visualize its commit and inspect its file:

Console images liberty

Alternatively, you can view the file by retrieving it from Pachyderm. Because this is an image, you cannot just print it out in the terminal, but the following command will let you view it:

On macOS, run:

pachctl get file images@master:liberty.png | open -f -a Preview.app

On Linux 64-bit, run:

pachctl get file images@master:liberty.png | display

Create a Pipeline #

Now that you have some data in your repo, it is time to do something with it. Pipelines are the core processing primitive in Pachyderm. Pipelines are defined with a simple JSON file called a pipeline specification or pipeline spec for short. For this example, we already created the pipeline spec for you.

When you want to create your own pipeline specification later, you can refer to the full Pipeline Specification to use more advanced options. Options include building your own code into a container. In this tutorial, your pipeline will use a pre-built Docker image.

📖

More about the concept of Pipeline in Pachyderm.

For now, we are going to create a single pipeline that takes in images and does some simple edge detection.

Below is the edges.json pipeline spec. Let’s walk through the details.

{
  "pipeline": {
    "name": "edges"
  },
  "description": "A pipeline that performs image edge detection by using the OpenCV library.",
  "transform": {
    "cmd": [ "python3", "/edges.py" ],
    "image": "pachyderm/opencv:1.0"
  },
  "input": {
    "pfs": {
      "repo": "images",
      "glob": "/*"
    }
  }
}

The pipeline spec contains a few simple sections. The pipeline section contains a name, which is how you will identify your pipeline. Your pipeline will also automatically create an output repo with the same name. The transform section allows you to specify the docker image you want to use. In this case, pachyderm/opencv:1.0 is the docker image (defaults to DockerHub as the registry), and the entry point is edges.py. The input section specifies repos visible to the running pipeline, and how to process the data from the repos. Commits to these repos will automatically trigger the pipeline to create new jobs to process them. In this case, images is the repo, and /* is the glob pattern.

The glob pattern defines how the input data will be transformed into datums if you want to distribute computation. /* means that each file can be processed individually, which makes sense for images. Glob patterns are one of the most powerful features in Pachyderm.

📖

More about the concept of Glob Pattern in Pachyderm and the fundamental notion of Datums.

The following extract is the Python code run in this pipeline:

import cv2
import numpy as np
from matplotlib import pyplot as plt
import os

# make_edges reads an image from /pfs/images and outputs the result of running
# edge detection on that image to /pfs/out. Note that /pfs/images and
# /pfs/out are special directories that Pachyderm injects into the container.
def make_edges(image):
   img = cv2.imread(image)
   tail = os.path.split(image)[1]
   edges = cv2.Canny(img,100,200)
   plt.imsave(os.path.join("/pfs/out", os.path.splitext(tail)[0]+'.png'), edges, cmap = 'gray')

# walk /pfs/images and call make_edges on every file found
for dirpath, dirs, files in os.walk("/pfs/images"):
   for file in files:
       make_edges(os.path.join(dirpath, file))

The code simply walks over all the images in /pfs/images, performs edge detection, and writes the result to /pfs/out.

/pfs/images and /pfs/out are special local directories that Pachyderm creates within the container automatically. All the input data for a pipeline is stored in /pfs/<input_repo_name> and your code should always write out to /pfs/out (see the function make_edges(image) above). Pachyderm automatically gathers everything you write to /pfs/out, versions it as this pipeline output, and maps it to the appropriate output repo of your pipeline.

Now, let’s create the pipeline in Pachyderm:

pachctl create pipeline -f https://raw.githubusercontent.com/pachyderm/pachyderm/2.4.x/examples/opencv/edges.json

Again, check the end result in your Console:

Console edges pipeline

What Happens When You Create a Pipeline #

Creating a pipeline tells Pachyderm to run your code on the data in your input repo (the HEAD commit) as well as all future commits that occur after the pipeline is created. Our repo already had a commit, so Pachyderm automatically launched a job to process that data.

📖

More about the concept of Job in Pachyderm.

The first time Pachyderm runs a pipeline job, it needs to download the Docker image (specified in the pipeline spec) from the specified Docker registry (DockerHub in this case). This first run might take a minute or so because of the image download, depending on your Internet connection. Subsequent runs will be much faster.

You can view the job with:

pachctl list job

System response:

ID                               SUBJOBS PROGRESS CREATED       MODIFIED
23378d899d3d45738f55df3809841145 1       ▇▇▇▇▇▇▇▇ 5 seconds ago 5 seconds ago

You can check the state of your pipeline:
```
pachctl list pipeline
```
System response:
```
NAME  VERSION INPUT     CREATED       STATE / LAST JOB  DESCRIPTION
edges 1       images:/* 2 minutes ago running / success A pipeline that performs image edge detection by using the OpenCV library.
```
Yay! Your pipeline succeeded! Pachyderm creates a corresponding output repo for every pipeline. This output repo will have the same name as the pipeline, and all the results of that pipeline will be versioned in this output repo. In our example, the edges pipeline created an output repo called edges to store the results written to /pfs/out.

List your repositories:

pachctl list repo

System response:

NAME   CREATED        SIZE (MASTER) ACCESS LEVEL
edges  10 minutes ago ≤ 22.22KiB    [repoOwner]  Output repo for pipeline edges.
images 3 hours ago    ≤ 57.27KiB    [repoOwner]

Note that all of that information and more are available in your Console.

Reading the Output #

We can view the output data from the edges repo in the same fashion that we viewed the input data.

On macOS, run:

pachctl get file edges@master:liberty.png | open -f -a Preview.app

On Linux 64-bit, run:

pachctl get file edges@master:liberty.png | display

The output should look like this:

Console edges liberty

Processing More Data #

Pipelines will automatically process the data from new commits as they are created. In a way, pipelines “subscribe” to their input repo(s), ready to process any new incoming commits. Also similar to Git, commits have a parental structure that tracks which files have changed. In this case we are going to be adding more images.

Let’s create two new commits in a parental structure. To do this we will simply do two more put file commands and by specifying master as the branch, it automatically parents our commits onto each other. Branch names are just references to a particular HEAD commit.

pachctl put file images@master:AT-AT.png -f http://imgur.com/8MN9Kg0.png
pachctl put file images@master:kitten.png -f http://imgur.com/g2QnNqa.png

Adding a new commit of data will automatically trigger the pipeline to run on the new data we have added. We will see corresponding jobs get started and commits to the output edges repo. Let’s also view our new outputs.

View the list of jobs that have started:

pachctl list job

System response:

ID                               SUBJOBS PROGRESS CREATED        MODIFIED
1c1a9d7d36944eabb4f6f14ebca25bf1 1       ▇▇▇▇▇▇▇▇ 31 seconds ago 31 seconds ago
fe5c4f70ac4347fd9c5934f0a9c44651 1       ▇▇▇▇▇▇▇▇ 47 seconds ago 47 seconds ago
23378d899d3d45738f55df3809841145 1       ▇▇▇▇▇▇▇▇ 12 minutes ago 12 minutes ago

View the output data:

On macOS, run:

pachctl get file edges@master:AT-AT.png | open -f -a Preview.app
pachctl get file edges@master:kitten.png | open -f -a Preview.app

On Linux, run:

pachctl get file edges@master:AT-AT.png | display
pachctl get file edges@master:kitten.png | display

Adding Another Pipeline #

We have successfully deployed and used a single stage Pachyderm pipeline. Now, let’s add a processing stage to illustrate a multi-stage Pachyderm pipeline (also referenced as a Directed Acyclic Graph or DAG is this documentation). Specifically, let’s add a montage pipeline that take our original and edge detected images and arranges them into a single montage of images:

Below is the pipeline spec for this new pipeline:

{
  "pipeline": {
    "name": "montage"
  },
  "description": "A pipeline that combines images from the `images` and `edges` repositories into a montage.",
  "input": {
    "cross": [ {
      "pfs": {
        "glob": "/",
        "repo": "images"
      }
    },
    {
      "pfs": {
        "glob": "/",
        "repo": "edges"
      }
    } ]
  },
  "transform": {
    "cmd": [ "sh" ],
    "image": "v4tech/imagemagick",
    "stdin": [ "montage -shadow -background SkyBlue -geometry 300x300+2+2 $(find /pfs -type f | sort) /pfs/out/montage.png" ]
  }
}

This montage pipeline spec is similar to our edges pipeline except for the following differences:

We are using a different Docker image that has imagemagick installed.
We are executing a sh command with stdin instead of a python script in the pipeline’s transform section.
We have multiple input data repositories (images and edges).

In the montage pipeline we are combining our multiple input data repositories using a cross pattern. This cross pattern creates a single pairing of our input images with our edge detected images. There are several interesting ways to combine data in Pachyderm, which are discussed in pipelines’ concepts and our pipeline specification page.

To create the montage pipeline, run:

pachctl create pipeline -f https://raw.githubusercontent.com/pachyderm/pachyderm/2.4.x/examples/opencv/montage.json

See your new DAG in Console:

Console opencv DAG

The pipeline creation triggers a job that generates a montage for all the current HEAD commits of the input repos:

pachctl list job

System response:

ID                               SUBJOBS PROGRESS CREATED        MODIFIED
01e0c8040e18429daf7f67ce34c3a5d7 1       ▇▇▇▇▇▇▇▇ 11 seconds ago 11 seconds ago
1c1a9d7d36944eabb4f6f14ebca25bf1 1       ▇▇▇▇▇▇▇▇ 12 minutes ago 12 minutes ago
fe5c4f70ac4347fd9c5934f0a9c44651 1       ▇▇▇▇▇▇▇▇ 12 minutes ago 12 minutes ago
23378d899d3d45738f55df3809841145 1       ▇▇▇▇▇▇▇▇ 24 minutes ago 24 minutes ago

View the generated montage image in Console or by running one of the following commands:

In Console:

Console opencv montage

On macOS, run:

pachctl get file montage@master:montage.png | open -f -a Preview.app

On Linux 64-bit, run:

pachctl get file montage@master:montage.png | display

Next Steps #

You can use what you have learned to build on or change these pipelines. You can also dig in and learn more details about:

Again, we would love to help and see what you come up with! Submit any questions, comment, contribution on GitHub, Slack, or email at support@pachyderm.io if you want to show off anything nifty you have created!