Skip to content

Latest commit

 

History

History
579 lines (393 loc) · 26.3 KB

README.md

File metadata and controls

579 lines (393 loc) · 26.3 KB
Raw

Kafka Tutorials

The source code for the Kafka Tutorials microsite. Read about it in our blog post.

Setup

If you want to hack on this site to add a new tutorial or make a change, follow these instructions.

Prerequisites

The following prerequisites are only required if you are going to run the microsite locally. If you are interested in testing tutorials locally see the Testing Locally section of the README.

  • ruby 2.3 or later
  • bundler
  • npm
  • python3 / pip3
  • gradle
  • docker-compose

On the Mac, you can get the dependencies like this:

brew install ruby node
gem install bundler

You'll now have an executable called harness-runner on your path. (Note that if you use Python, you likely already have the pyyaml package installed.)

Installing

1. Clone this repository

git clone [email protected]:confluentinc/kafka-tutorials.git

Then cd into the directory.

2. Install the node packages

npm install

This will bring in some external JavaScript and CSS packages that we're using.

3. Install the gems

bundle install

This will install Jekyll itself and any other gems that we use.

4. Run the development server

bundle exec jekyll serve --livereload

This will launch a web server so that you can work on the site locally. Check it out on http://localhost:4000.

5. Install the Pip package

This repository uses a Python package to facilitate testing the tutorials. To keep things simple, we bundled it into this repository. You can get everything you need by running the following:

pip3 install pyyaml
cd harness_runner
pip3 install -e .

6. Running the tests locally

You can run any of the automated tests locally using a command like:

cd ..
make -C _includes/tutorials/filtering/ksql/code tutorial

Substituting in the appropriate tutorial path.

Note: if you are testing unreleased ksqlDB Docker images, you will need to be logged in to AWS ECR.

Add code for a new tutorial

A tutorial is a short procedure, targeted at developers, for getting a certain thing done using Confluent Platform.

In many cases, you can get that thing done using one of several stacks. For example, you might be able to perform data filtering by writing a KSQL query, by writing a Kafka Streams application, or by directly using the Kafka Consumer API. These comprise the three stacks this site supports: ksql, kstreams, and kafka.

Kafka Tutorials is a bit unique in that each tutorial is self-testing. That is, we have built a light-weight harness system that's able to instrument the code that belongs to each tutorial to make sure that it actually works. This is really useful as we expect to have a lot of tutorials.

With in each stack, these tutorials contain a few pieces. These are described below.

Using tutorial author tools

There are several pieces that you will put together for a tutorial. Aside from the original content you need to provide (tutorial description, actual tutorial content, code, etc.), all of the steps described below can be automated. You accomplish tutorial automation via one of two scripts in the tools directory

  1. gen_project.sh
  2. clone_tutorial.sh

Let's describe each script.

The gen_project.sh script

As the name implies, the gen_project.sh script generates the minimal structure for a viable Kafka Tutorials tutorial. The code and text contained in the tutorial are place holders, and you'll update those with your code and writing to complete the tutorial.

The script expects you to supply a properties file for setting the name and other parts of the tutorial. Here's an example properties file you can use.

#!/bin/sh

# space-separated list of types could be one of ksql or kstreams
CARDS="kstreams ksql"

# all lower-case with no spaces
TUTORIAL_SHORT_NAME=my-tutorial-name

# the MAIN_CLASS variable will generate a test named MAIN_CLASSTest
MAIN_CLASS=FilteringCode
AK_VERSION=2.7.0
CP_VERSION=6.1.0
KSQLDB_VERSION=0.11.0
SEMAPHORE_TEST_NAME="Test name for Semaphore run"
PERMALINK="seo-friendly-link-to-my-tutorial"

To run this script:

  1. Create a branch
  2. Make sure you are in the kafka-tutorials base directory
  3. Execute ./tools/gen_project.sh my_tutorial_props.sh

You'll see a lot of information scroll across the screen, describing each step of the tutorial generation process. The last part of the information presented is a checklist of what you'll need to do to complete your tutorial, aside from adding your code and tutorial text.

Your tutorial, my-tutorial-name, has been generated successfully!
The script adds a copy of this checklist to the my-tutorial-name directory.

There are some additional steps you'll need to take to complete the tutorial:


1. Update the following entries in _data/tutorials.yaml file
    a. title
    b. meta-description
    c. question
    c. introduction

You can find these fields by searching for my-tutorial-name in the _data/tutorials.yaml file.


2. Update the link text then add the following link(s) to the index.html file in the appropriate section:


   <li><a href="seo-friendly-link-to-my-tutorial/kstreams.html">MEANINGFUL LINK TEXT HERE</a></li>

   <li><a href="seo-friendly-link-to-my-tutorial/ksql.html">MEANINGFUL LINK TEXT HERE</a></li>

If you only specified one tutorial type (ksql or kstreams) then you'd only have one link in the output. Also, a copy of this output is copied in your tutorial directory /Kafka Tutorials base dir/_includes/tutorials/you_tutorial_short_name

The clone-tutorial.sh script

As the name implies, this script creates a clone of an existing tutorial. The clone script changes the name of the tutorial throughout the content. To clone a tutorial, you also need to provide a properties file

#!/bin/sh

#SAMPLE-PROPERTIES-FILE-CONTENT

# By using the basename of a tutorial everything is cloned possibly ksql, kstreams, and kafka 
ORIG_TUTORIAL=connect-add-key-to-source

# to only clone a kafka tutorial
ORIG_TUTORIAL=console-consumer-producer-basic/kafka

# to clone just the ksql part
ORIG_TUTORIAL=connect-add-key-to-source/ksql

# to clone just the kstreams portion
ORIG_TUTORIAL=connect-add-key-to-source/kstreams

!!!! IMPORTANT YOU MUST ONLY HAVE ONE ORIG_TUTORIAL VARIABLE !!!!

# Add a new tutorial name
NEW_TUTORIAL=junkA
# Add a new sempahore test name
SEMAPHORE_TEST_NAME="My New Junk Tutorial"
# Add a new PERMALINK
PERMALINK=a-junk-tutorial

To clone a tutorial:

  1. Create a branch
  2. Make sure you are in the kafka-tutorials base directory
  3. Execute ./tools/clone_tutorial.sh my-clone-tutorial-props.sh

You'll see a similar output scroll across the screen, including the checklist for items you'll need to do for a completed tutorial.

Which script to use?

How do you decide which script to run? If you are creating a new tutorial that does not resemble an existing tutorial, then the gen_project script is probably the better way to go. If you are creating a tutorial that is closely related to a current tutorial, a hopping-windows tutorial, when there is already a tumbling-windows tutorial, for example, then the clone approach is probably better.

It's still valuable to read through the next section to learn how all the tutorial pieces fit together.

Description of tutorial parts

If you used the clone script then many of these will already exist and will just need customising for your particular tutorial.

1. Describe the question your tutorial answers

The first thing to do is articulate what question your tutorial is meant to answer. Every tutorial contains a question and an example scenario. Edit _data/tutorials.yml and add your entry. The top item in this file represents the short name for your tutorial. For example, the tutorial for transforming events of a stream is transforming. You'll also notice a status attribute. You can enable as many stacks as you'd like to author for this tutorial, but we recommend starting with just one.

2. Make the directory structure

Next, make a few directories to set up for the project:

mkdir _includes/tutorials/<your tutorial short name>/<stack>/code
mkdir _includes/tutorials/<your tutorial short name>/<stack>/markup

3. Write the code for the tutorial

Add your code for the tutorial under the code/ directory you created. This should be entirely self-contained and executable with a docker-compose.yml file and a platform-appropriate build. Follow the conventions of existing tutorials as closely as possible.

At this point, you should feel free to submit a PR! A member of Confluent will take care of writing the markup and test files to integrate your code into the site. You can, of course, proceed to the next section and do it yourself, if you'd like.

Add a narrative and test for the tutorial

This section is generally for those who work at Confluent and will be integrating new tutorials into the site. We need to do a little more work than just authoring the code. We also need to write the markup to describe the tutorial in narrative form, and also write the tests that we described to make sure it all works. This section describes how to do that.

1. Create a harness for the tutorial

The harness is the main data structure that allows us to both test and render a tutorial from one form. Make a new directory under _data/harnesses/ for your tutorial slug name and stack, like _data/harnesses/<your tutorial short name>/ksql.yml. Follow the existing harnesses to get a feel for what this looks like. The main thing to notice is that each step has a render attribute that points to a file. Create the markup for this in the next section.

2. Create markup for the tutorial

Under the markup/ directory that was created earlier, create 3 subdirectories: dev, test, and prod. Write the tutorial prose content here, following the conventions of existing tutorials. These files should be authored in Asciidoc.

3. Tie it all together

Make a file named /tutorials/<your tutorial short name>/<stack>.html, specifying all the variables of interest. Note: the directory structure for these files is distinct from /_includes/tutorials.

For example, to display the tutorial with the ksqlDB stack:

# /tutorials/filter-a-stream-of-events/ksql.html
---
layout: tutorial
permalink: /tutorials/filter-a-stream-of-events/ksql
stack: ksql
static_data: filtering
---

You can do the same for Kafka Streams and Kafka, by using the kstreams and kafka stacks, respectively.

4. Add your tutorial into build system

Lastly, create a Makefile in the code directory to invoke the harness runner and check any outputs that it produces. Then modify the .semaphore/semaphore.yml file to invoke that Makefile. This will make sure your tutorial gets checked by the CI system.

Testing Locally

The Kafka Tutorials microsite shows how to manually execute each tutorial step-by-step. However, there are some scenarios when a user may want to run and test a tutorial in a more automated fashion:

  • End-to-end: user makes a small change to the code and wants to validate that the tutorial still works end-to-end
  • Run-and-play: user runs a tutorial and wants to leave it running to play with the environment

This section describes how you can do either of these scenarios using the harness-runner to programmatically run a single tutorial.

Prerequisites

The following prerequisites are required if you are going to run a tutorial programmatically.

  • python3 / pip3
  • gradle
  • docker-compose

Environment Setup

  1. Check out the kafka-tutorials GitHub repo:
git clone https://github.com/confluentinc/kafka-tutorials.git
cd kafka-tutorials
  1. Install the packages for the harness runner.

If you have pip3 installed locally:

(cd /harness_runner/ && pip3 install -e .)

If you don't have pip3 installed locally, create a Dockerfile with the following content:

FROM python:3.7-slim
RUN pip3 install pyyaml

and then run the following command to build and execute the Docker image:

docker build -t runner . ; docker run -v ${PWD}/harness_runner:/harness_runner/ -it --rm runner bash -c 'cd /harness_runner/ && pip3 install -e .'

  1. Install gradle for tutorials that compile any code.

  2. Install Docker Compose

Run a tutorial

  1. (optional) If you want to augment or override a tutorial's Docker environment, set the Docker Compose CLI environment variable COMPOSE_FILE to include docker-compose.yml and the absolute path to a docker-compose.override.yml file. For example, to use Confluent Control Center with any Kafka Tutorial, set COMPOSE_FILE to docker-compose.yml and the absolute path to this docker-compose.override.yml.
export COMPOSE_FILE=docker-compose.yml:<path to Kafka Tutorials>/tools/docker-compose.override.yml
  1. End-to-end: execute the harness runner for a single tutorial by calling make, across all dev, test, and prod stages, to validate it works end-to-end. Identify the tutorial you want and then run make. Note that this destroys all the resources and Docker containers it created, so it cleans up after itself. Format: (cd _includes/tutorials/<tutorial name>/<type>/code && make) where type is one of ksql | kstreams | kafka. Example:
(cd _includes/tutorials/transforming/kstreams/code/ && make)
  1. Run-and-play: execute the harness runner for a single tutorial by calling make SEQUENCE='"dev, test"', just across dev and test stages, which leaves all resources and Docker containers running so you can then play with it. Format: (cd _includes/tutorials/<tutorial name>/<type>/code && make SEQUENCE='"dev, test"') where type is one of ksql | kstreams | kafka. Example:
(cd _includes/tutorials/transforming/kstreams/code/ && make SEQUENCE='"dev, test"')

Now you can play with the environment, some sample commands shown below.

docker exec broker kafka-topics --list --bootstrap-server localhost:9092
docker exec -it ksqldb-cli ksql http://ksqldb-server:8088

Because the Docker containers are left running, don't forget to clean up when you are done.

docker container ls
docker container rm -f <container id>

Makefile Details

The Makefile will delete and re-create the outputs directory used to contain files with output from various steps used to verify the tutorial steps.

Here is the contents of an actual Makefile :

STEPS_DIR := tutorial-steps
DEV_OUTPUTS_DIR := $(STEPS_DIR)/dev/outputs
TEMP_DIR := $(shell mktemp -d)

tutorial:
  rm -r $(DEV_OUTPUTS_DIR) || true
  mkdir $(DEV_OUTPUTS_DIR)
  harness-runner ../../../../../_data/harnesses/fk-joins/kstreams.yml $(TEMP_DIR)
  diff --strip-trailing-cr $(STEPS_DIR)/dev/expected-output-events.json $(DEV_OUTPUTS_DIR)/music-interest.json

The last line uses the diff command to validate expected output against the tutorial steps' actual output. The Makefile may have more than one validation action to have separate diff commands for each verification.

For example, here's the Makefile from the dynamic output topics tutorial

STEPS_DIR := tutorial-steps
DEV_OUTPUTS_DIR := $(STEPS_DIR)/dev/outputs
TEMP_DIR := $(shell mktemp -d)

tutorial:
  rm -r $(DEV_OUTPUTS_DIR) || true
  mkdir $(DEV_OUTPUTS_DIR)
  harness-runner ../../../../../_data/harnesses/dynamic-output-topic/kstreams.yml $(TEMP_DIR)
  diff --strip-trailing-cr $(STEPS_DIR)/dev/expected-output.json $(DEV_OUTPUTS_DIR)/actual-output.json
  diff --strip-trailing-cr $(STEPS_DIR)/dev/expected-special-output.json $(DEV_OUTPUTS_DIR)/actual-special-order-output.json

Harness Details

Given that the test harness is the heart of a tutorial, it will be helpful to describe in detail how to work with a kafka|ksql|kstreams.yml file. You should note the harness file is in the YAML file format, so formatting properly is essential. The harness files generate the rendered tutorial structure and validate any output of tutorial steps against expected values.

New tutorial authors should not need to create a harness file from scratch, using either the tools/gen_project.sh or tools/clone.sh script will provide a usable harness file. This section should provide enough guidance to add, update, or remove parts as needed.

1. Structure

Three top-level sections make up the harness file:

  • dev - the setup and teaching portion of the tutorial (required)
  • test - test setup and execution of tests, if any (optional)
  • prod - steps to build and deploy a docker image of the tutorial code (optional)

In some cases, having a test and/or prod section doesn't make sense, so you can omit those portions of the harness file. The Apache Kafka console producer and consumer basic operations and the Apache Kafka console consumer with primitive keys and values tutorials are an excellent example of tutorials that don't need a test or prod section.

The dev, test, and prod sections contain a top-level element steps. The steps contains any number of well, steps for the user to walk through. Additionally, the harness_runner script follows the same steps for executing the tutorial automatically during builds. All sections contain the same step structure, so we'll only discuss the make-up of a single section.

For reference here is an example section of the harness file from the console consumer primitive keys and values tutorial

 - title: Get Confluent Platform
      content:
        - action: make_file
          file: docker-compose.yml
          render:
            file: tutorials/console-consumer-primitive-keys-values/kafka/markup/dev/make-docker-compose.adoc

        - action: execute_async
          file: tutorial-steps/dev/docker-compose-up.sh
          render:
            file: tutorials/console-consumer-primitive-keys-values/kafka/markup/dev/start-compose.adoc

        - action: execute
          file: tutorial-steps/dev/wait-for-containers.sh
          render:
            skip: true

        - name: wait for ksqldb-server and connectors
          action: sleep
          ms: 30000
          render:
            skip: true

  • Title - each section starts with a title element, and as the label suggests, the text provided here is the text used the label the step for the tutorial user and the output to the console by the harness runner. The title section contains one element - content

  • content - the content section (a YAML dictionary) contains an arbitrary sized list of YAML dictionaries named action. An action key creates an anonymous step, i.e., not specified in the test runner's output. For output in the test runner, you can provide a -name key, followed by some text for console output.

  • action - action keys drive the behavior of the harness. An action key can be one of these values

    • action: make_file - Prompts the user to create a file for the tutorial.
    • action: execute - a synchronous action step
    • action: execute_async - an asynchronous step, this indicates a step the user will keep running for some portion of the tutorial.
    • action: sleep - pause the test runner for an amount of time specified by the ms key. You use sleep key to allow some async action to complete
      • ms: NNN - the time in milliseconds you want the test harness to pause execution. You only use ms after an action: sleep entry.
    • docker_ksql_cli_session - an action starting a ksqlDB CLI session for working through a tutorial

In the next sections, you'll see how to use action keys to organize your harness files.

2. Action type descriptions and examples

  • make_file The make_file instructs the user to create a file an existing file required to run the tutorial. Some examples are the docker-compose.yml file, a statements.sql (ksqlDB), and Java files.

    A make_file with a file to render look like this:

    - action: make_file
      file: docker-compose.yml
      render:
        file: tutorials/console-consumer-primitive-keys-values/kafka/markup/dev/make-docker-compose.adoc
    • file: the path and name to the file. The path is relative to <tutorial-name>/<type>/code> .The harness runner will use the file during the automated tutorial execution.
    • render: the file to render. The render key has one of the two possible keys:
      • file: the path and name of the file to render to the user.

  • execute The execute step is a synchronous execution step during the tutorial for the user and the harness runner

     - action: execute
      file: tutorial-steps/dev/init.sh
      render:
        file: tutorials/console-consumer-primitive-keys-values/kafka/markup/dev/init.adoc

    Sometimes the execute step is an internal step only for the harness runner. Internal steps ignore the render: using a skip: true entry

      - name: wait for ksqldb-server and connectors
      action: sleep
      ms: 30000
      render:
        skip: true

    The execute steps may also capture output from stdout for tests executed by the harness runner

    - title: Invoke the tests
  content:
    - action: execute
      file: tutorial-steps/test/run-tests.sh
      stdout: tutorial-steps/test/outputs/test-results.log
      render:
        file: tutorials/aggregating-sum/ksql/markup/test/run-tests.adoc

    Here you can see the stdout: key specifying the file used to capture the output of an execute step.

  • execute_async The execute_async step is for steps needing to execute in the background while the user continues going through the tutorial

    Here's an example of running docker-compose to start docker containers to run the duration of the tutorial

     - action: execute_async
       file: tutorial-steps/dev/docker-compose-up.sh
       render:
         file: tutorials/console-consumer-primitive-keys-values/kafka/markup/dev/start-compose.adoc

    Here's another example of using a console consumer. Note the use of stdout: to capture the output from the consumer for testing.

     - title: Start an initial console consumer
    content:
      - action: execute_async
        file: tutorial-steps/dev/harness-console-consumer-keys.sh
        stdout: tutorial-steps/dev/outputs/actual-output-step-one.txt
        render:
          file: tutorials/console-consumer-primitive-keys-values/kafka/markup/dev/consume-topic-no-deserializers.adoc

  • docker_ksql_cli_session The docker_ksql_cli_session sets up ksqlDB tutorial users to start a ksqlDB CLI session so the user can execute various SQL files to complete the tutorial.

      - action: docker_ksql_cli_session
      container: ksqldb-cli
      docker_bootup_file: tutorial-steps/dev/start-cli.sh
      column_width: 20
      render:
        file: tutorials/aggregating-sum/ksql/markup/dev/start-cli.adoc

    The docker_ksql_cli_session contains the following keys:

    • container: - The name of the ksqldb-cli docker image in the docker-compose.yml file.
    • docker_bootup_file: - Tutorial users and the harness runner execute this file to start the dockerized CLI session.
    • column_width: - Formats the ksqlDB query output
    • render: - The harness renders the file corresponding to the file: key to tutorial users with the command to start the CLI session.
    • stdin: - stdin key contains one or more file keys specifying the different SQL file to execute for the tutorial.
        stdin:
        - file: tutorial-steps/dev/create-movie-ticket-sales.sql
          render:
            file: tutorials/aggregating-sum/ksql/markup/dev/create-movie-ticket-sales.adoc

        - file: tutorial-steps/dev/populate-movie-ticket-sales.sql
          render:
            file: tutorials/aggregating-sum/ksql/markup/dev/populate-movie-ticket-sales.adoc
    • stdout: - Contains directory: specifying the directory to capture all query output.

Updating dependency versions

The following regular expressions may be useful to group-update all dependencies within the repo:

  • confluentinc\/cp-([^:]*):\d+.\d+.\d+ will match all Confluent Platform components, except ksqlDB. Capture group 1 can be used to build the component name.
  • confluentinc\/ksqldb-([^:]*):\d+.\d+.\d+ will match all ksqlDB components. Capture group 1 can be used to build the component name.

Updating kafka-tutorials.confluent.io

The release branch tracks the content and code comprising the live site. Confluent manages the release process.

The ksqldb-latest branch builds against the latest master branch of ksqlDB, and should be used for updates that are only in the master branch of ksqlDB. Confluent manages the processes of merging changes from this branch.

Prepare a release PR

A pull request into the release branch denotes a request to update the live site. The PR description should summarize the content changes and link to the staging site with the updates. In general, releases are timed, so target dates should also be noted.

Release artifacts are automatically promoted to the live site by CI, as part of successful release branch builds.