Skip to content

Latest commit

 

History

History
379 lines (241 loc) · 25.5 KB

File metadata and controls

379 lines (241 loc) · 25.5 KB

Kyma - Build, Pack and Push your Docker Images

  • Kyma
  • Cloud Foundry

Important - This part of the tutorial is required for Kyma deployments only!

As you might know, Kubernetes and therefore also Kyma is all about containerization of workloads. Our application is no exception from this pattern, so before you can deploy the SaaS sample application to your Kyma Cluster using Helm, you first need to build and containerize the various components (UI modules, CAP applications, Application Router aso.) by packing them into a Docker Image and pushing this Docker Image to a Container Registry of your choice like e.g., DockerHub. This is what the following chapter will be all about!

Let us have a brief look at the tool prerequisites, which are essential to build and push your containerized components.

If you are facing any issues during the following steps of our tutorial, please feel free to consult the excellent Developer Tutorial on Deploy Your CAP Application on SAP BTP Kyma Runtime. It describes similar steps and will get you covered in great detail, in case you get stuck in our sample scenario.

https://developers.sap.com/mission.btp-deploy-cap-kyma.html

1. Prerequisites

First of all, you will need Node.js installed on your device. You can find the latest LTS (long-term support) version (here). To conduct the upcoming steps, you also need the following tools installed on your device. Follow the linked instructions based on your system environment.

  • Docker (https://docs.docker.com/get-docker/) - To follow this tutorial on Kyma, you will require an application that allows you to handle container images on your computer, including functions like building, pushing, pulling, and running them.
  • Pack CLI (https://buildpacks.io/docs/install-pack/) - By utilizing Pack, you can construct container images based on collectively maintained base images, resulting in a more straightforward process for maintaining and updating them.

    Installation Shortcuts
    MacOS - brew install buildpacks/tap/pack
    Windows - choco install pack

Important - Please make sure to comply with the respective license agreements, especially when using Docker in an enterprise context!

Before you continue, make sure you can successfully execute the following commands in your command line. Make sure to create a new command line instance for testing the commands. This ensures Docker and the Pack CLI have been added to your Environment Variables. In case of errors, please make sure that you followed all steps of the instructions provided.

> docker
A self-sufficient runtime for containers
...
> pack
CLI for building apps using Cloud Native Buildpacks
...

Furthermore, please make sure you have access to a Container Registry (e.g. DockerHub) of your choice. You can use the following Docker CLI command to login to your Container Registry. Please consult the documentation of your preferred Container Registry provider to learn more about login requirements.

Hint - If you do not have access to a Container Registry yet, please create a new (free) account e.g., on DockerHub (click here) or choose a hyperscaler Container Registry by AWS, Azure or GCP. Alternatively, you can also use GitHub to store and manage your Docker images (learn more).

Important - Please be aware of the fact, that free service providers like DockerHub, only allow you to store a limited number of private Docker Images. If your Docker Containers contain sensible information (which they shouldn't) or intellectual property, make sure to choose a service plan or provider that allows you to create private Docker Images for your purpose!

> docker login
...
Login Succeeded

Once you successfully installed the required tools (Docker & Pack) and logged in to your Container Registry, you can continue with the next step.

2. Build your components

Let us first build the various components of our SaaS sample application, before containerizing them into Docker Images.

2.1. (Fork and) Clone our sample GitHub repository to your development environment using the below commands. Switch to the deploy/kyma directory which contains all objects required for a Kyma deployment.

git clone https://github.com/SAP-samples/btp-cap-multitenant-saas
cd deploy/kyma

2.2. Run an initial npm install. This will install all required dependencies for the build processes and also for local testing in the code directory.

## Run in ./deploy/kyma ## 
npm install --include=dev --prefix=../../code

Alternatively, you can also run the following npm script.

## Run in ./deploy/kyma ## 
npm run install

2.3. Execute the cds build --production command to trigger a new production build of the available CAP components (Shared Data Model, Tenant Data Model, Backend Service, API Service). You can change the version of the package by attaching it with an additional "@"-sign like "@sap/[email protected]".

## Run in ./deploy/kyma ## 
npx -p @sap/cds-dk@8 cds build -in ../../code --profile production

Alternatively, you can also run the following npm script.

## Run in ./deploy/kyma ## 
npm run build

2.3. Build the available UI5 modules by running the following npm command. This will take a few minutes for the first build. Consecutive will be faster.

Hint - This build process, will use the official UI5 Tools to build your applications and copies them into the code/app/html5-deployer/resources directory for deployment.

## Run in ./deploy/kyma ## 
npm run ui:apps

This is it! The application components requiring a separate build step (CAP & UI5) have been catered, and can now be containerized into Container Images.

3. Create your Docker Images

While in Cloud Foundry, there is no requirement to containerize your application components before deployment, in Kyma or respectively Kubernetes, you need to take care of this yourself. While this sounds like a complicated effort, we tried to simplify things as much as we can for you!

Containerization is required, as during the deployment to your Cluster, Helm will pull the Docker Images of our the sample application components. This includes Docker Image for the following application components:

  • API Service (Pod)
  • API Service Broker (Pod)
  • Application Router (Pod)
  • Backend Service (Pod)
  • HTML5 Apps Deployer (Job)
  • HDI Container Deployer (Job)

These components will be containerized in Docker Images in the following steps. To simplify this process, we are relying on SAP Standard Docker Images or so-called Cloud Native Buildpacks. Using Cloud Native Buildpacks, saves you major effort compared to creating and maintaining your own Dockerfiles. Instead you will always use latest images provided by the Cloud Native Buildpacks project, which is also part of the Cloud Native Computing Foundation. If you are new to topics like Kubernetes, we strongly recommend to check out both pages to learn more!

Important - To learn more about how the Docker Images for the various components are being build, read the next part of this chapter.

3.1. Run the following npm script (from within your deploy/kyma directory), which will build all Docker Images using SAP Standard Docker Images and Cloud Native Buildpacks.

Hint - If you use e.g. DockerHub as a Container Registry, please put in your username (e.g., johndoe) as Container Image Prefix placeholder. If you use the GitHub Container Registry, the prefix will look similar to ghcr.io/<namespace> (e.g. ghcr.io/johndoe). All generated Docker Images will be automatically prefixed with this label!

Hint - Using devices with ARM chips (e.g., Apple M1) the build process involving Cloud Native Buildpacks might take several minutes. Please do not immediately cancel the build if things appear to be stuck, but wait some time for the process to continue (especially while the SBOM is being generated)!

## Run in ./deploy/kyma ## 
npx cross-env IMAGE_PREFIX=<ContainerImagePrefix> npm run build:all

# Example
npx cross-env IMAGE_PREFIX=sap-demo npm run build:all

Alternatively, you can also build the Docker Images separately by running the component specific npm scripts. Check the package.json file to find all available scripts.

## Run in ./deploy/kyma ## 
npx cross-env IMAGE_PREFIX=<ContainerImagePrefix> npm run build:srv

# Example
npx cross-env IMAGE_PREFIX=sap-demo npm run build:srv

Once the process is finished, your Docker Images are ready to be pushed to the Container Registry of your choice. Before doing this, let us briefly check what has happened under the hood to build your Docker Images.

4. Build details

As explained, our sample application uses different build tools and approaches to create the required Docker Images, which are Paketo backed by Cloud Native Buildpacks and official SAP Standard Docker Images. This part of the tutorial will provide you some high-level insights on the Docker Image build process of the different components. While it might cause some confusion on first sight, why actually two different approaches are being used to build the required Docker Images, the explanation is fairly simple.

Building Dockerfiles for Kubernetes workloads is not always easy. Especially for beginners in this context! Also the maintenance of Dockerfiles can be time consuming. So whenever we can skip that responsibility and use a vendor-maintained Docker Image (like e.g., Standard Docker Images provided by SAP), we make use of it. For all other components, where SAP is not providing any Standard Docker Images or scenarios where it does not make any sense to do so (e.g., for custom Node.js apps), we make use of Paketo and Cloud Native Buildpacks.

Once again, I highly suggest to have a quick break and scroll through the excellent blog post of Maximilian Streifeneder, who gets you covered on the topic of Paketo as well as some nice little tricks how to further analyze generated Docker Images!

https://blogs.sap.com/2023/03/07/surviving-and-thriving-with-the-sap-cloud-application-programming-model-deployment-to-sap-btp-kyma-runtime/

Docker Images created using Paketo and Cloud Native Buildpacks are secure, efficient, production-ready and come with a lot of features, which are hard to provided using Dockerfiles and would require much more manual effort. To get an idea of the features provided by Cloud Native Buildpacks, check out the official documentation (click here). To learn about the general concepts behind Cloud Native Buildpacks (turning your source-code into a read-to-use Docker Image), check out the respective documentation (click here).

So much for the introduction. Let us now check, how the various components of our SaaS sample application are being containerized either using Paketo and Cloud Native Buildpacks or existing SAP Standard Docker Images.

API Service (api)

Build using Paketo and Cloud Native Buildpacks

As the API Service is based on CAP and Node.js, for the initial build or any changes, you first need to run a cds build --production. After running the CDS build (compiling the CDS files of our CAP Backend, CAP API, Tenant and shared data model), the API Service Docker Image is build using Paketo and Cloud Native Buildpacks.

This simplifies the containerization process and allows you to build a Docker Image without the necessity of maintaining a separate Dockerfile for Node.js workloads. During the build process, Paketo will take the content of the gen/api directory and put it into the working directory of a Node.js Docker Image. This image is based on the latest and stable Cloud Native Buildpacks.

Npm script to build the Docker Image using Paketo (/deploy/kyma/package.json)

"build:api": "pack build sap-demo/susaas-api --path ../../code/gen/api --builder paketobuildpacks/builder:base --buildpack gcr.io/paketo-buildpacks/nodejs -e BP_LAUNCHPOINT=./node_modules/@sap/cds/bin/cds-serve.js"

Hint - The paketobuildpacks/builder:base provides a minimal Docker Image based on Ubuntu, with the addition of a few packages (so-called "mixins") (click here for more details).

Backend Service (srv)

Build using Paketo and Cloud Native Buildpacks

As the central Backend Service is also based on CAP, for the initial build or any changes, you first need to run a cds build --production. After running the CDS build (compiling the CDS files of our CAP Backend, CAP API, Tenant and shared data model), also for the Backend Service, Paketo and Cloud Native Buildpacks are used to create a Docker Image.

Doing so (as for the API Service), a Docker Image can be build without having to maintain a separate Dockerfile. During the build process, Paketo will take the content of the gen/srv directory, and place it into the working directory a Node.js Docker Image. Again, this image is based on the latest and stable Cloud Native Buildpacks.

Npm script to build the Docker Image using Paketo (/deploy/kyma/package.json)

"build:srv": "pack build sap-demo/susaas-srv --path ../../code/gen/srv --builder paketobuildpacks/builder:base --buildpack gcr.io/paketo-buildpacks/nodejs -e BP_LAUNCHPOINT=./node_modules/@sap/cds/bin/cds-serve.js"

Application Router (router)

Build using SAP-provided Standard Docker Image

The Application Router Docker Image is based on SAP's standard sapse/approuter Docker Image which can be found in Docker Hub (https://hub.docker.com/u/sapse). As this official Docker Image is optimized and maintained by SAP, there is no need to make use of Cloud Native Buildpacks for a stable, secure and up-to-date base images.

Therefore, we created a tiny Dockerfile which is using the official SAP Docker Image (sapse/approuter), and only adds our own static Application Router resources to the working directory of the Docker Image. In this case, the custom Dockerfile is required to copy the whole content of the router directory, except for the package.json (part of .dockerignore file) to the resulting Docker Image.

Hint - The package.json file is part of the directory for local testing purposes only. As the SAP-managed Docker Image sapse/approuter already contains a package.json file, we will reuse the start script of this SAP-provided package.json scripts.

Dockerfile based on sapse/approuter Docker Image (/code/router/Dockerfile)

# Image based on SAP provided sapse/approuter image
# Make sure to update the Docker Image version regularly
FROM sapse/approuter:x.x.x

# Create an app directory (in case not there yet)
RUN mkdir -p /app
# Grant default non-root user ("node") permissions
RUN chown node.node /app

# Set working directory for COPY and CMD command
WORKDIR /app

# Copy local directory where Dockerfile resides to Docker Image
# -> Includes the resources folder and the xs-app.json file
COPY . .
# Run npm start script to start the Application Router
CMD [ "npm", "start" ]

Build Docker Image based on Dockerfile above (/deploy/kyma/package.json)

Hint - This following npm script will build a Docker Image based on the Dockerfile located in the /code/router directory and will tag it with sap-demo/susaas-router. "sap-demo" has to be replaced by the prefix also used in step 3.1, if you want to run this command standalone.

"build:router": "docker build -t sap-demo/susaas-router ../../code/router"

API Service Broker (broker)

Build using Paketo and Cloud Native Buildpacks

For the API Service Broker (which is actually just a generic Node.js workload based on the @sap/sbf npm module - click here for details), again Paketo and Cloud Native Buildpacks are used to build the required Docker Image.

This way (as for the API and SaaS Backend Service), a Docker Image can be build without maintaining a separate Dockerfile. During the build process, Paketo will take the content of the broker directory, and place it into the working directory of a Node.js Docker Image. This Docker Image is again based on the latest Cloud Native Buildpacks.

Npm script to build the Docker Image using Paketo (/deploy/kyma/package.json)

"build:broker": "pack build sap-demo/susaas-broker --path ../../code/broker --builder paketobuildpacks/builder:base --buildpack gcr.io/paketo-buildpacks/nodejs -e BP_LAUNCHPOINT=./start.js"

HTML5 Apps Deployer (app/html5-deployer)

Build using SAP-provided Standard Docker Image

The HTML5 Apps Deployer Docker Image of our sample is (like the Application Router) based on SAP's official sapse/html5-app-deployer Docker Image which can be found in Docker Hub (https://hub.docker.com/u/sapse). The purpose of the HTML5 Apps Deployer is actually pretty simple. The HTML5 Apps Deployer is based on an npm package (click here for details), which will take our zipped UI5 modules and deploys them into the HTML5 App Repository Service, which will be bound to the respective Kyma workload.

As this Docker Image is maintained by SAP, there is no need to make use of Cloud Native Buildpacks for stable, secure and up-to-date base image. Therefore, we created a Dockerfile which is based on the official SAP Docker Image (sapse/html5-app-deployer), and only add our own resources folder to the working directory of the resulting Docker Image.

The resources folder (within the code/app/html5-deployer directory) contains the zipped UI5 modules (which you need to build upfront using the npm run ui:apps script - see here). If you followed all previous tutorial steps, the respective zip files should already exist, as the respective command will automatically run the build scripts in the package.json files of our UI5 modules.

Npm script to trigger the build of a single UI module (/deploy/kyma/package.json)

"ui:admin-projects": "npm run build:copy --prefix ../../code/app/ui-admin-projects/"

Npm scripts to build a UI module and copy zip to html5 app deployer (/code/app/ui-admin-projects/package.json)

"scripts": {
    "build:copy": "npm run build && npm run copy",
    "build": "ui5 build preload --clean-dest --config ui5-deploy.yaml --include-task=generateCachebusterInfo",
    "copy": "shx mkdir -p ../html5-deployer/resources/ && shx cp -rf ./dist/*.zip ../html5-deployer/resources/"
}

Similar to the Application Router, the Dockerfile residing in the code/app/html5-deployer folder, copies the html5-deployer directory content into the working directory of the resulting Docker Image, which is based on the SAP-managed sapse/html5-app-deployer image.

Hint - The package.json file is part of the code/app/html5-deployer directory for local testing purposes only. As the Docker Base Image sapse/html5-app-deployer already contains a corresponding package.json file, we will reuse the start script of this SAP-provided package.json.

Dockerfile based on sapse/html5-app-deployer Docker Image (/code/app/html5-deployer/Dockerfile)

# Image based on SAP provided sapse/html5-app-deployer image
# Make sure to update the Docker Image version regularly
FROM sapse/html5-app-deployer:x.x.x

# Create the app directory (in case not there yet)
RUN mkdir -p /app
# Grant default non-root user ("node") permissions
RUN chown node.node /app

# Set the working directory for COPY and CMD command
WORKDIR /app

# Copy the directory where Dockerfile resides to the Docker Image
# -> Includes the resources directory with the zipped UI5 modules
COPY . .
# Run npm start script to start the Application Router
CMD [ "npm", "start" ]

Build Docker Image based on Dockerfile above (/deploy/kyma/package.json)

"build:html5-deployer": "docker build -t sap-demo/susaas-html5-deployer ../../code/app/html5-deployer"

HDI Container Deployer (db-com)

Build using Paketo and Cloud Native Buildpacks

As the data model of the shared database container is also based on CAP, for the initial build or any changes, you first need to run cds build --production. While the Tenant specific data model will be automatically deployed to isolated HDI containers upon subscription, the shared data model has to be deployed (into a dedicated HDI Container) separately. Therefore, the default @sap/hdi-deploy npm package (click here for details) is being used by this Docker Image.

After running the CDS build (compiling the CDS files of our CAP Backend, CAP API, Tenant and shared data model), the required Docker Image (containing an HDI deployer for the shared data model) is build using Paketo and Cloud Native Buildpacks. This simplifies the build process and allows us to build a Docker Image without the necessity of maintaining a separate Dockerfile. During the build process, Paketo will take the content of the gen/db-com directory and place it into the working directory of a Node.js Docker Image. This image is based on the latest Cloud Native Buildpacks.

Npm script to build the Docker Image using Paketo (/deploy/kyma/package.json)

"build:db-com": "pack build sap-demo/susaas-db-com --path ../../code/gen/db-com --builder paketobuildpacks/builder:base --buildpack gcr.io/paketo-buildpacks/nodejs -e BP_LAUNCHPOINT=./node_modules/@sap/hdi-deploy/deploy.js"

Now you should be covered and well in the know about the different approaches used to containerize our SaaS application components. Let us continue with some more details on how to push your Docker Images to your Container Registry now.

5. Push your Docker Images

To allow Helm to pull your Docker Images during the deployment process, you need to push them to a Container Registry of your choice.

Hint - If you are using a non-public Docker Registry, we assume you know how to handle the required secrets or access codes. In these scenarios, you might need to adapt the respective npm scripts in the package.json file.

After all your Docker Images are build using Cloud Native Buildpacks or SAP Standard Images, you can push them to your Container Registry. To do so, please ensure you successfully logged in to your registry of choice (docker login) before running the following npm script.

Hint - If you use e.g. DockerHub as a Container Registry, please put in your username (e.g., johndoe) as Container Image Prefix placeholder. If you use the GitHub Container Registry, the prefix will look similar to ghcr.io/<namespace> (e.g. ghcr.io/johndoe). All generated Docker Images will be automatically prefixed with this label!

## Run in ./deploy/kyma ## 
npx cross-env IMAGE_PREFIX=<ContainerImagePrefix> npm run push:all

# Example
npx cross-env IMAGE_PREFIX=sap-demo npm run push:all

As an alternative to pushing all Docker Images at once, you can again push images separately by running the component specific npm script.

Hint - This can be useful if you e.g., updated only a single Docker Image and will save you some time.

## Run in ./deploy/kyma ## 
npx cross-env IMAGE_PREFIX=<ContainerImagePrefix> npm run push:srv

# Example
npx cross-env IMAGE_PREFIX=sap-demo npm run push:srv

Well, this is it! Once the npm script has finished, your Docker Images should be available in your Container Registry. You have successfully containerized all application components can start deploying the SaaS sample application to your Kyma Cluster. Let's move on to the actual deployment!

6. Further information

Please use the following links to find further information on the topics above.