Skip to content

Latest commit

 

History

History
404 lines (263 loc) · 15 KB

readme-local-building-example.md

File metadata and controls

404 lines (263 loc) · 15 KB
Raw

Local building example

Introduction

The Docker Hub has removed the auto-building feature from the free plan since 2021-07-26.

This page describes how to build the images locally and optionally also push them to the Docker Hub.

The second version (G3v2) of the building pipeline makes it really easy, even if you want to build a set of images or all of them at once.

Preparation

Ensure file attributes after cloning

It may be necessary to repair the executable files attributes after cloning the repository (by git clone).

You can do that by executing the following commands from the project's root directory:

find . -type f -name "*.sh" -exec chmod +x '{}' \;
chmod +x docker/hooks/*

For example, if the files in the folder docker/hooks would not be executable, then you would get errors similar to this:

$ ./builder.sh latest build

==> EXECUTING @2023-03-05_16-42-57: ./builder.sh 

./builder.sh: line 84: ./docker/hooks/build: Permission denied

Set environment variables before building

Open a terminal windows and change the current directory to the root of the project (where the license file is).

Make a copy of the secrets example file, modify it and then source it in the terminal:

### make a copy and then modify it
cp examples/example-secrets.rc secrets.rc

### source the secrets
source ./secrets.rc

### or also

. ./secrets.rc

TIP: If you copy a file named secrets.rc into the folder docker/hooks/, then it will be automatically sourced by the hook script env.rc.

Be aware that the following environment variables are mandatory and must be always set:

  • REPO_OWNER_NAME
  • BUILDER_REPO

Ensure that your secrets.rc file contains at least the lines similar to these:

export REPO_OWNER_NAME="accetto"
export BUILDER_REPO="headless-drawing-g3"

You can use your own names if you wish.

Alternatively you can modify the hook script file env.rc like this:

### original lines
declare _owner="${REPO_OWNER_NAME:?Need repo owner name}"
DOCKER_REPO="${_owner}/${BUILDER_REPO:?Need builder repo name}"

### modified lines
declare _owner="${REPO_OWNER_NAME:-accetto}"
DOCKER_REPO="${_owner}/${BUILDER_REPO:-headless-drawing-g3}"

Again, you can use your own names if you wish.

You can also use other ways to set the variables.

Ensure wget utility

If you are on Windows, you can encounter the problem of missing wget utility. It is used by refreshing the g3-cache and it's available on Linux by default.

On Windows you have generally two choices. You can build your images inside the WSL environment or you can download the wget.exe application for Windows. Make sure to update also the PATH environment variable appropriately.

Building pipeline

The actual building pipeline consists of the following hook scripts stored in the folder docker/hooks:

  • pre_build
  • build
  • push
  • post_push

The hook scripts are executed exactly in that order.

The second version (G3v2) of the pipeline has added also the helper script cache, which ist stored in the same folder. It is used by the hook scripts pre_build and build and it refreshes the local g3-cache. It can be also executed stand-alone.

Utilizing the local g3-cache brings a significant boost in the building performance and much shorter building times.

There is also the helper script util-readme.sh, stored in the folder utils/. This script can be used for preparing the README file for the Docker Hub.

Three ways of building images

Since the second version (G3v2) of the building pipeline there are the following ways of building the images:

  • Building sets of images by executing the helper script ci-builder.sh
  • Building the individual images by executing the helper script builder.sh
  • Building the individual images by executing the hook scripts step-by-step

Remark: You can force the building of a new persistent image in any case by setting the environment variable FORCE_BUILDING=1.

Building and publishing sets of images

Building and publishing of sets of images is pretty easy. Let's say that we want to refresh the images that feature the Blender. We can do that by executing the following command:

### PWD = project's root directory
./ci-builder.sh all group blender blender-chromium blender-firefox

### or also
./ci-builder.sh all group complete-blender

### or skipping the publishing to the Docker Hub
./ci-builder.sh all-no-push group complete-blender

The script ci-builder.sh is using the utility script builder.sh internally.

You can find more information and examples in the separate readme file, describing the utility ci-builder.sh.

Remark: The set can contain also a single image.

Building and publishing individual images

Building and publishing of individual images is also very easy. Let's say we wan to refresh the image accetto/ubuntu-vnc-xfce-opengl-g3:latest. We could execute the following command:

### PWD = project's root directory
### 'latest' stands for 'opengl' in this context
./builder.sh latest all

### alternatively with additional Docker CLI options
./builder.sh latest all --no-cache

### or skipping the publishing to the Docker Hub
./builder.sh latest all-no-push

The script builder.sh is using the individual hook scripts internally.

You can find more information and examples in the separate readme file, describing the utility builder.sh.

Step-by-step building and publishing

The building pipeline can executed also step-by-step. The hook scripts in the folder docker/hooks/ can be executed directly or also by using the utility script builder.sh.

The script builder.sh is using the individual hook scripts internally.

Step 1: pre_build

### PWD = project's root directory

./builder.sh latest pre_build
### or
./docker/hooks/pre_build dev latest

### alternatively with additional Docker CLI options
./builder.sh latest pre_build --no-cache
### or
./docker/hooks/pre_build dev latest --no-cache

This step builds the temporary helper image and creates the following temporary helper files that are used by other scripts:

  • scrap-version_sticker_current.tmp
  • scrap-version_sticker-verbose_current.tmp
  • scrap-version_sticker-verbose_previous.tmp
  • scrap-demand-stop-building

The file scrap-demand-stop-building is created only if the verbose version sticker hasn't changed since the last time it has been published on the builder repository's GitHub Gist and if the environment variable FORCE_BUILDING is not set to 1. Its presence will block the next hook script build from building a new persistent image. If you want to force the image building, you can delete this file manually.

The other option is to set the environment variable FORCE_BUILDING=1 before executing the pre_build script.

Remark: The temporary helper image is automatically removed by the script.

Step 2: build

### PWD = project's root directory

./builder.sh latest build
### or
./docker/hooks/build dev latest

### alternatively with additional Docker CLI options
./builder.sh latest build --no-cache
### or
./docker/hooks/build dev latest --no-cache

This step builds a new persistent image, named by default according the variable BUILDER_REPO (e.g. headless-drawing-g3).

Remark: Ensure that the file scrap-demand-stop-building is not present or the environment variable is set FORCE_BUILDING=1.

Step 3: push

### PWD = project's root directory

./builder.sh latest push
### or
./docker/hooks/push dev latest

This step creates the deployment image tag and pushes it to the deployment repository on the Docker Hub.

Step 4: post_push

### PWD = project's root directory

./builder.sh latest post_push
### or
./docker/hooks/post_push dev latest

This step updates the GitHub Gists belonging to the builder repository and the deployment repository and removes the temporary helper files created by the hook script pre_build.

Additional parameters

Additional parameters, that come after the mandatory ones, can be passed to the hook scripts in the folder docker/hooks.

The individual hook scripts can use the additional parameters or ignore them.

Currently only the build and pre_build scripts actually process the additional parameters.

Both scripts insert the additional parameters just after the docker build part of the Docker build command line.

For example, if the additional parameters --target stage_xfce --no-cache are provided to the script docker/hooks/build, then the result Docker command line will begin like this:

docker build --target stage_xfce --no-cache -f ./docker/Dockerfile.xfce.22-04 ...

However, there is a special handling of the parameter --target.

Special handling of --target parameter

The Docker build parameter --target allows processing multi-stage Dockerfiles only to a particular stage.

Therefore this parameter makes sense only by the hook script docker/hooks/build.

For example, this would build an image including only the stages stage_essentials, stage_xserver and stage_xfce:

docker build --target stage_xfce --no-cache -f ./docker/Dockerfile.xfce.22-04 ...

The result image tag will get the suffix _stage_xfce, e.g. accetto/headless-ubuntu-g3:latest_stage_xfce.

This is generally useful only during development and debugging.

On the other hand, the hook script docker/hooks/pre_build ignores and removes the parameter --target with its value.

Therefore pre_build always process all Dockerfile stages.

For example, if the additional parameters --target stage_xfce --no-cache are provided to the script docker/hooks/pre_build, then the result Docker command line will begin like this:

docker build --no-cache -f ./docker/Dockerfile.xfce.22-04 ...

Note that both hook scripts always remove an orphaned --target parameter which comes with no value.

Disabling default features

Some features, that are enabled by default, can be disabled via environment variables.

It allows to build even smaller images by excluding noVNC or Firefox Plus features.

Disabling noVNC

If the environment variable FEATURES_NOVNC is explicitly set to zero (by export FEATURES_NOVNC="0"), then

  • image will not include noVNC
  • image tag will get the -vnc suffix (e.g. latest-vnc etc.)

Disabling Firefox Plus

If the environment variable FEATURES_FIREFOX_PLUS is explicitly set to zero (by export FEATURES_FIREFOX_PLUS="0") and the variable FEATURES_FIREFOX="1", then

  • image with Firefox will not include the Firefox Plus features
  • image tag will get the -default suffix (e.g. latest-firefox-default or also latest-firefox-default-vnc etc.)

README files for Docker Hub

Each deployment repository has its own README file for the Docker Hub.

The source README file for the Docker Hub is split into two parts.

The first part is called README-dockerhub.md and it is the actual content of the README file for the Docker Hub.

The second part is called readme-append.template and it contains the templates for the badge links.

The final README file for the Docker Hub can be produced by the utility script util-readme.sh, which is stored in the utils/ folder.

The result file will be named scrap-readme.md and its content can be copy-and-pasted manually to the README area of the repository on the Docker Hub.

The utility will also check the length of the result file, because it is limited to max. 25000 characters by the Docker Hub.

Therefore there is always also the full-length README file version, which is published only on the GitHub.

For example, the README file for the repository accetto/ubuntu-vnc-xfce-opengl-g3 can be generated by the following command:

### PWD = utils/
./util-readme.sh --repo accetto/ubuntu-vnc-xfce-opengl-g3 --context=../docker/xfce --gist <deployment-gist-ID> -- preview

### or if the environment variable 'DEPLOY_GIST_ID' has been set
./util-readme.sh --repo accetto/ubuntu-vnc-xfce-opengl-g3 --context=../docker/xfce -- preview

Tips and examples

How to deploy all images into one repository

There are six deployment repositories by default.

Their names are defined by the following environment variables:

  • DEPLOYMENT_REPO_BLENDER for Blender images
  • DEPLOYMENT_REPO_DRAWIO for drawio-desktop images
  • DEPLOYMENT_REPO_FREECAD for FreeCAD images
  • DEPLOYMENT_REPO_GIMP for GIMP images
  • DEPLOYMENT_REPO_INKSCAPE for Inkscape images
  • DEPLOYMENT_REPO_OPENGL for generic images with Mesa3D/OpenGL/VirtualGL

The forth variable BUILDER_REPO defines the name of the builder repository, which is not used for deployment by default.

However, the images can be optionally published also into the builder repository by setting the environment variable FORCE_PUBLISHING_BUILDER_REPO=1.

The images in the builder repository are distinguished by their tags.

On the other hand, if the repository environment variables are unset or set to the reserved value void, then the deployment into the related repository will be skipped.

This behaviour can be used, if you want to publish all the images into a single repository.

Simply set all repositories except the builder one to void and force the publishing into the builder repository.

For example, for publishing all the images into a single repository headless-drawing-g3 set the variables like this:

DEPLOYMENT_REPO_BLENDER="void"
DEPLOYMENT_REPO_DRAWIO="void"
DEPLOYMENT_REPO_FREECAD="void"
DEPLOYMENT_REPO_GIMP="void"
DEPLOYMENT_REPO_INKSCAPE="void"
DEPLOYMENT_REPO_OPENGL="void"
BUILDER_REPO="headless-drawing-g3"
FORCE_PUBLISHING_BUILDER_REPO=1