Skip to content

Latest commit

 

History

History
948 lines (727 loc) · 31.1 KB

README.md

File metadata and controls

948 lines (727 loc) · 31.1 KB
Raw

ISLE: Docker Prototype

LICENSE Build and Push Docker Images

Introduction

This repository provides a number of docker images which can be used to build an Islandora site. On commit, these images are automatically pushed to Docker Hub via Github Actions. Which are consumed by isle-dc and isle-site-template. They can also be used by other Docker orchestration tools such as Swarm / Kubernetes. Reach out on the community slack for other example installations.

It is not meant as a starting point for new users or those unfamiliar with Docker, or basic server administration.

If you are looking to use islandora please read the official documentation and use either isle-dc or the isle-site-template to deploy via Docker or the islandora-playbook to deploy via Ansible.

Requirements

To build the Docker images using the provided Gradle build scripts requires:

N.B You can use older versions of Docker to run the images, just not build them.

N.B If you use Firefox you will also have to install nss to use mkcert see the docs.

N.B The version of make that comes with OSX is to old, please update using brew etc.

To verify you have all the requirements run the following command.

make setup

If it is unsuccessful you should see the following message:

Could not find executable: XXXX
Consult the README.md for how to install requirements.

Windows

Since this repository relies on make you must use WSL2 to invoke it, and all dependencies should be installed inside of the WSL2 context. With the exeception of mkcert which must be installed in the host Windows context, as it needs to modify the Windows trust store so that the browsers will support the generated certificates.

N.B Firefox is not a supported browser on Windows for this project, you must use Edge or Chrome.

Tooling

There are a number of tools you can use to build and test the images produced by this repository. In general there are tools like docker buildx and docker compose that can be invoked directly or you can the wrapper tools like make, gradle. Using the wrapper tools has some advantages and is generally recommended, but it is occasionally good to revert to the tools they wrap around if you need to debug an issue with the building or testing.

Make

Building and running relies on Make. You can see the tasks available by invoking Make with no arguments. Make is really only supported in Linux/OSX, or if you are using Windows you can use Windows Subsystem for Linux (WSL)

make

Which yields:

Usage:
  make <target>

BuildKit:
  bake                           Builds and loads the target(s) into the local docker context.
  push                           Builds and pushes the target(s) into remote repository.
  manifest                       Creates manifest for multi-arch images.

Compose:
  up                             Starts up the local development environment.
  down                           Stops the local development environment.

General:
  test                           Run all tests
  clean                          Destroys local environment and cleans up any uncommitted files.
  purge                          Destroys all data.
  help                           Displays this help message.

Gradle

Testing and generating security reports as well as DockerHub maintenance rely on Gradle and should function equally well across platforms. The only difference being the script you call to interact with gradle (the following assumes you are executing from the root directory of the project):

Linux or OSX:

./gradlew

Windows:

gradlew.bat

For the remaining examples the Linux or OSX call method will be used, if using Windows substitute the call to Gradle script.

Gradle is a project/task based build system to query all the available tasks use the following command.

./gradlew tasks --all

Which should return something akin to:

> Task :tasks

------------------------------------------------------------
Tasks runnable from root project
------------------------------------------------------------

...
Isle DockerHub tasks
--------------------
deleteEligibleDockerHubTags - Delete eligible tags from DockerHub 'islandora/cache' Repository.
getDockerHubTagsEligibleForDeletion - Gets the tags eligible for removal from DockerHub 'islandora/cache' Repository.
getDockerHubToken - Gets the login token required for interacting with DockerHub Rest API.
getProtectedDockerHubTags - Gets the tags which should not be removed by DockerHub cleanup inactive tags task.

Isle Reports tasks
------------------
grype - Process the software bill of material with Grype
pullGrype - Pull anchore/grype docker image
pullSyft - Pull anchore/syft docker image
syft - Generate a software bill of material with Syft
updateGrypeDB - Update the Grype Database

Isle Tests tasks
----------------
cleanUpAfter - Clean up resources after running test
cleanUpBefore - Clean up resources before running test (if interrupted externally, etc)
setUp - Prepare to run test
test - Perform test

...

In Gradle each Project maps onto a folder in the file system path where it is delimited by : instead of / (Unix) or \ (Windows).

The root project : can be omitted.

So if you want to run a particular task taskname that resided in the project folder project/subproject you would specify it like so:

./gradlew :project:subproject:taskname

To get more verbose output from Gradle use the --info argument like so:

./gradlew :PROJECT:TASK --info

Github Actions

This repository makes use of Github Actions to perform a number of tasks.

Workflow Description
cleanup.yml Deletes old tags in DockerHub once a week.
dockerhub-description.yml Updates DockerHub Description of images to match README.md files.
push.yml Builds and Tests images and generates a security vulnerability report.

Building

This repository makes use of buildx which is a wrapper around buildkit to build all the images. This can be invoked directly like so:

docker buildx bake <target>

By default if no <target> is specified it will build all the images in this repository.

Alternatively you can use make to invoke bake, it is advantageous as it will pass in additional properties that should allow you to make use of the remote cache.

make bake

If you want to build cross platform issues, please use the isle-builder repository for setting up a builder.

Build All Images

The following will build all the images in the correct order.

make bake

By default this will produce images like islandora/base:local you can change the repository and tag by specifying them as arguments, like so:

make bake REPOSITORY=foo TAGS=bar

This would produce images like foo/base:bar.

Build Specific Image

To build a specific image and it's dependencies, for example islandora/tomcat, you can use the following:

make bake TARGET=tomcat

If you do not want to build all it's dependent images you can set the context using existing images like so:

make bake TARGET=nginx CONTEXTS="docker-image://islandora/base:1.0.10"

N.B. Github actions do this to prevent rebuilding dependant images.

Testing

There are a number of automated tests that are included in this repository which can be found in the tests folders of each docker image project.

To run these tests use the following command:

./gradlew test

N.B. Running all tests concurrently can saturate Docker's default number of bridge networks. Please see the Issues/FAQ for how to remedy this.

Test Specific Image

Alternatively you can test a single image like so:

./gradlew tomcat:test

Running

While isle-buildkit does provide a test environment, it is not meant for development on Islandora or as production environment. It is meant for testing for breaking changes to the images provided by this repository. Instead please refer to isle-dc, or the isle-site-template, for how to build your own Islandora site.

To manually test changes in a functioning environment you can use the provided docker-compose.yml file.

Though you must first generate certificates for use by traefik.

make certs

N.B. This will prompt you for a password as generating a root Certificate requires administrative privileges.

You can interact with the docker compose directly instead of using make.

docker compose up -d

Although you can interact with docker compose directly, it is recommend you use make as this will ensure you have build all the images and generated the required certificates needed, etc:

make up

This will bring up the environment based on islandora-starter-site. When completed a message will print like so:

Waiting for installation...

  Credentials:
  Username                       admin
  Password                       password

  Services Available:
  Drupal                         https://islandora.dev
  IDE                            https://ide.islandora.dev
  ActiveMQ                       https://activemq.islandora.dev
  Blazegraph                     https://blazegraph.islandora.dev/bigdata/
  Fedora                         https://fcrepo.islandora.dev/fcrepo/rest/
  Matomo                         https://islandora.dev/matomo/index.php
  Solr                           https://solr.islandora.dev
  Traefik                        https://traefik.islandora.dev

To stop the containers use the following command:

make stop

To destroy this environment use the following command:

make down

The two commands can be used at once to ensure you are starting from a clean environment:

make down up

IDE

An IDE is provided at http://ide.islandora.dev which includes:

  • Intellisense & Code Completion
  • Build tasks for performing lints & other common actions (accessible via CTRL-B or CMD-B)
  • Integrated Debugger (XDebug)
  • PHPCS / PHPCBF
  • etc...

To enable XDebug when using drush via the built-in terminal enter the following command before invoking drush:

export XDEBUG_SESSION=1

For web requests, you must also send an XDEBUG_SESSION cookie with your request, this can be toggled on and off via a browser plugin such as the following.

PHPStorm

PHPStorm and alternative IDE's which allow for remote development via SSH are also supported.

Add the following to your ~/.ssh/config file:

Host islandora.dev
  ForwardAgent yes
  PasswordAuthentication yes
  Port 2222
  PreferredAuthentications password
  PubkeyAuthentication no
  StrictHostKeyChecking no
  User nginx
  UserKnownHostsFile /dev/null

You should now be able to ssh like so (assuming you've already brought the docker compose environment up):

ssh islandora.dev

You can then connect via the PHP remote development feature.

PHPStorm

N.B. PHPStorm remote is not supported form Arm architectures so the above will not work on M1 macbooks and later.

Docker Images

The following docker images are provided:

Many are intermediate images used to build other images in the list, for example java. Please see the README.md of each image to find out what settings, and ports, are exposed and what functionality it provides, as well as how to update it to the latest releases.

Additionally this repository consumes imagemagick & leptonica images produced by a separate repositories. Since it is a standalone image that rarely changes and takes a while to build, due to building it under emulation.

Updating Dependencies

To update the dependencies of a image follow this general pattern, for example alpaca.

Update the following ARG values in

ARG ALPACA_VERSION="x.x.x"
ARG ALPACA_FILE_SHA256="xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"

You'll have to download the new version you wish to update to, you can construct the url from the the following ARG values in the Dockerfile, to generate the sha256 value to put in the above ARG.

ARG ALPACA_VERSION="x.x.x"
ARG ALPACA_FILE="islandora-alpaca-app-${ALPACA_VERSION}-all.jar"
ARG ALPACA_URL="https://repo1.maven.org/maven2/ca/islandora/alpaca/islandora-alpaca-app/${ALPACA_VERSION}/${ALPACA_FILE}"

For example for the version x.x.x:

ALPACA_VERSION="x.x.x"
ALPACA_FILE="islandora-alpaca-app-${ALPACA_VERSION}-all.jar"
ALPACA_URL="https://repo1.maven.org/maven2/ca/islandora/alpaca/islandora-alpaca-app/${ALPACA_VERSION}/${ALPACA_FILE}"
wget "${ALPACA_URL}" &>/dev/null
shasum -a 256 ${ALPACA_FILE}

N.B. Please read the release notes the new version and account for any changes to configuration that are required, as well as test locally.

Updating Composer

A number of images like crayfish provide a composer.lock file to pin them to particular dependencies.

When updating the dependencies be sure to search for composer.lock in the rootfs folder of the image and update them as well.

This can be done by running the image after updating dependencies, to get the latest code, and running composer update. For example crayfish:

# Update ARGS as done in previous section
# ...
# Build image
make bake TARGET=crayfish
for lock in $(find crayfish -name "composer.lock"); \
do \
  docker run --rm -ti -v "$(pwd)/${lock}:${lock#crayfish/rootfs*}" -w $(dirname "${lock#crayfish/rootfs*}") --entrypoint composer islandora/crayfish:local update; \
done

Updating Configuration

In particular review the files in rootfs/etc/confd, as configuration is likely to change between releases.

Design Considerations

All of the images build by this project are derived from the Alpine Docker Image which is a Linux distribution built around musl libc and BusyBox.

N.B. While musl libc is of general higher quality vs. glibc, it is less commonly used and many libraries have come to depend on the undefined behavior of glibc so in some of our images we patch in glibc to ensure their correct function.

The image is only 5MB in size and has access to a package repository. It has been chosen for its small size, and ease of generating custom packages (as is done in the imagemagick & leptonica images).

The base image includes two tools essential to the functioning of all the images.

  • Confd: Configuration Management
  • S6 Overlay: Process Manager / Initialization system

Confd

confd is used for all Configuration Management, it is how images are customized on startup and during runtime. For each Docker image there will be a folder rootfs/etc/confd that has the following layout:

./rootfs/etc/confd
├── conf.d
│   └── file.ext.toml
└── templates
    └── file.ext.tmpl

The file.ext.toml and file.ext.tmpl work as a pair. The toml file defines where the template will be render to and who owns it. The tmpl file being the template in question. Ideally these files should match the same name of the file they are generating minus the toml or tmpl suffix. This is to make their discovery easier.

Additionally in the base image there is confd.toml which sets defaults such a the log-level:

backend = "env"
confdir = "/etc/confd"
log-level = "error"
interval = 600
noop = false

confd is also the source of all truth when it comes to configuration. We have established a order of precedence in which environment variables at runtime are defined.

  1. Confd backend (highest)
  2. Secrets kept in /run/secrets (Except when using Kubernetes)
  3. Environment variables passed into the container
  4. Environment variables defined in Dockerfile(s)
  5. Environment variables defined in the /etc/defaults directory (lowest only used for multiline variables, such as JWT)

If not defined in the highest level the next level applies and so forth down the list.

N.B. /etc/defaults and the environment variables declared in the Dockerfile(s) used to create the image are required to define all environment variables used by scripts and confd templates. If not specified in either of those locations the environment variables will not be available even if its defined at a higher level i.e. confd.

The logic which enforces these rules is performed in container-environment.sh

N.B Some containers derive environment variables dynamically from other environment variables. In these cases they are expected to provided an additional oneshot services that must be executed before the confd-oneshot so that the variables are defined before confd is used to render templates.

By either using the command with-contenv or starting a script with #!/command/with-contenv bash the environment defined will follow the order of precedence above. Additionally Within confd templates it is required to use getenv function for fetching data, as the final value is written to the container environment.

S6 Overlay

S6 Overlay is the process supervisor we use in all the containers. It ensures initialization happens in the correct order and services start in the correct order (e.g. fpm-php starts prior to nginx, etc).

There are two types of services:

  • oneshot Services: Short lived services, used to prepare the container prior to running services
  • longrun Services: Long lived services like Nginx

Both types of services can have dependencies on one another, which indicates the order in which they are executed. oneshot services are run to completion before their dependent services are executed. longrun services are meant to run indefinitely, if for some reason one fails the container will stop and exit with the code of the failed service (provided a finish script is provided).

The longrun services have the following structure:

./rootfs/etc/s6-overlay/s6-rc.d
└── SERVICE_NAME
    ├── dependencies.d
    │   └── base
    ├── finish
    ├── run
    └── type

The run script is responsible for starting the service in the foreground. The finish script can perform any cleanup necessary before stopping the service, but in general it is used to kill the container, like so:

/run/s6/basedir/bin/halt

To declare dependencies between services, just add an empty file with the services name in it's dependencies.d folder.

For scripts we want to run at startup run we must register them. This can be done by placing an empty file named for the service in ./rootfs/etc/s6-overlay/s6-rc.d/user/contents.d.

There are only a few longrun services:

  • activemq
  • confd (optional, not enabled by default)
  • fpm
  • mysqld
  • nginx
  • solr
  • tomcat
  • etc

Of these only confd can be configured to run in every container, it periodically listens for changes in it's configured backend (e.g. etcd or environment variables) and will re-render the templates upon any change. See it's README.md, for more information.

oneshot services are pretty much the same, except they use they up and down instead of run and finish.

Additionally up is an execline script and does not support bash. So we typically just call out to a bash script instead, which by convention can be found in ./rootfs/etc/s6-overlay/scripts.

One oneshot service is of particular interest to all the containers. The ready service, which does not do anything in and of itself. It is meant as a placeholder that other services can rely on to ensure that typical actions have been performed, such as the configuration of environment variables, the rendering of templates and so on.

N.B. All longrun services should have a dependency on the ready service.

If you need to wait until a service to be ready for use, use the following command:

# Wait for PHP-FPM to report it has started.
s6-svwait -U /run/service/fpm

N.B. This requires the service to make use of notification-fd, which at the time of writing is only implemented for nginx and php-fpm

Image Hierarchy

In order to save space and reduce the amount of duplication across images, they are arranged in a hierarchy, that roughly follows below:

└── base
    ├── java
    │   ├── activemq
    │   ├── alpaca
    │   ├── solr
    │   └── tomcat
    │       ├── blazegraph
    │       ├── cantaloupe
    │       ├── fcrepo6
    │       └── fits
    ├── mariadb
    ├── postgresql
    └── nginx
        ├── crayfish
        │   ├── homarus
        │   ├── houdini (consumes [imagemagick] as well during its build stage)
        │   ├── hypercube (consumes [leptonica] as well during its build stage)
        │   ├── milliner
        │   └── riprap
        ├── crayfits
        ├── drupal
        │   └── test
        └── matomo

imagemagick & leptonica stand outside of the hierarchy as they are use only to build packages that are consumed by other images during their build stage.

Folder Layout

To make reasoning about what files go where each image follows the same filesystem layout for copying files into the image.

A folder called rootfs maps directly onto the linux filesystem of the final image. So for example rootfs/etc/islandora/configs will be /etc/islandora/configs in the generated image.

Build System

Since bake is used to build all the images, you must add new images to docker-bake.hcl.

Be sure to update IMAGES and DEPENDENCIES variables for any new images added, along with all the required targets for your new IMAGE-NAME:

  • IMAGE-NAME-common: Properties shared by all the following targets.
  • IMAGE-NAME: Targets the host platform.
  • IMAGE-NAME-amd64: Targets amd64, regardless of host platform.
  • IMAGE-NAME-arm64: Targets arm64, regardless of host platform.
  • IMAGE-NAME-ci: Used to update the remote cache and build both PLATFORM-ci images.
  • IMAGE-NAME-amd64-ci: Targets amd64, regardless of host platform updates remote cache.
  • IMAGE-NAME-arm64-ci: Targets arm64, regardless of host platform updates remote cache.

Multi-arch builds

For a number of the following reasons we've split the builds to build a single image per architecture at a time rather than concurrently building multi-arch images with buildkit even though that is a supported feature.

  • Some Docker repositories such as GitLab's and Amazons do not handle OCI multi-arch images well, and it's good to be able to fall back onto a architecture specific image.
  • The buildkit caching system is bugging when dealing with concurrent builds of multi-arch systems. It is better to split the caches by architecture to ensure cache stability.
  • Being able to refer to an image by it's architecture directly is useful when testing for cross platform bugs.

That being said we still produce the OCI manifests for multi-arch images. For local development this is not a requirement but the Github Actions will do this when building.

make manifest TARGET=tomcat

So for example on any newish Docker the following command.

docker run --rm -ti --entrypoint uname islandora/base:latest -a

Will pull the appropriate image for the host platform without the user explicitly specifying which of the images to e.g.

docker run --rm -ti --entrypoint uname islandora/base:latest-amd64 -a

Where the above command will explicitly pull the amd64 image regardless of the hosts architecture.

N.B. By default local builds will not build multi-arch images, they will only build the platform supported by the host. The above really only is used by the Github Actions build jobs.

Caching

The caching provided by buildkit is somewhat finicky and hard to control. We've opted to use registry-cache by default. No other forms of caching are supported.

Additionally things like the s3 is too slow to be practical, and gha (Github Actions cache) is not large enough to support all the images we build.

Design Constraints

To be able to support a wide variety of backends for confd, as well as orchestration tools, all calls must use getenv for the default value. With the exception of keys that do not get used unless defined like DRUPAL_SITE_{SITE}_NAME. This means the whatever backend for configuration, wether it be etcd, consul, or environment variables, containers can successfully start without any other container present. Additionally it ensure that the order of precedence for configuration settings.

This does not completely remove dependencies between containers, for example, when the fcrepo6 starts it requires a running database like mariadb to be able to start. In these cases an oneshot service can block until another container is available or a timeout has been reached. For example:

# Need access to database to start wait up to 5 minutes (i.e 300 seconds).
if timeout 300 wait-for-open-port.sh "${DB_HOST}" "${DB_PORT}" ; then
    echo "Database Found"
else
    echo "Could not connect to Database"
    exit 1
fi

This allows container to start up in any order, and to be orchestrated by any tool.

Issues / FAQ

Question: I'm getting the following error when building:

failed to solve with frontend dockerfile.v0: failed to solve with frontend
gateway.v0: runc did not terminate successfully: context canceled

Answer: If possible upgrade Docker to the latest version, and switch to using the Overlay2 filesystem with Docker.

Question: I'm getting the following error when running many tests at once:

ERROR: could not find an available, non-overlapping IPv4 address pool among the
defaults to assign to the network

Answer: By default Docker only allows 31 concurrent bridge networks to be created, but you can change this in your /etc/docker/daemon.json file by adding the following, and restarting Docker:

{
  "default-address-pools" : [
    {
      "base" : "172.17.0.0/12",
      "size" : 20
    },
    {
      "base" : "192.168.0.0/16",
      "size" : 24
    }
  ]
}