Simplified Docker dev setup

[birkland]

Draft PR to demonstrate Islandora 8 in Docker, using a simplified setup containing an external Ansible variables file, and no special Docker inventory, in contrast to #124

Creates and runs two Docker containers:

• An ansible container that, upon startup, runs ansible to install Islandora 8 to the vm container
• A vm container that plays the role of an empty Centos 7 box upon which Islandora 8 is installed via the ansible container.

Motivation

• Providing the start of a development environment that would allow modeling of "services deployed on separate hosts" as containers.
  • Right now, though, everything is installed to a single container whose role is analogous to the Vagrant VM. It's a start, though.
• Because using Ansible + Vagrant + Virtualbox in Windows is possible, but impractically awkward

To Test

• Make sure Docker is installed, as well as docker-compose (usually comes with the Docker distribution, e.g on Mac and Windows)
• Make sure the Islandora ports are free on your machine (e.g. there isn't another Islandora running on the same ports using Vagrant)
• Check out this PR, and in the repository's directory (claw-playbook/) do docker-compose build
• Do docker-compose up -d
• Watch the logs of the ansible container via docker logs -f ansible. That's the ansible output!
• At this point, the only way to know it's done/ready is to look at the logs. So watch the logs until Ansible is finished
  • Once the Ansible stops, the container will just run infinitely tailing /dev/null. The container won't exit if it succeeds or fails. It runs infinitely because that has been helpful for me in debugging Ansible issues.
• If Ansible succeeds, try logging into Islandora at the usual address localhost:8000, and poke around!

Tips and tricks

• You can hop onto any container while it's running by docker exec -it /bin/bash. For example, docker exec -it vm /bin/bash
• docker-compose stop will stop all containers, but keep state. You can later on do a docker-compose up vm just to start the Islandora "vm" container.
  • I haven't yet put the persisted state into volumes, so files/state are just written in the container. As a result, all state disappears in a poof once the containers are destroyed, e.g. via docker-compose down. docker-compose stop preserves state.

[whikloj]

Trying this out as the normal method is failing and I have a problem with the ansible box.

Step 6/8 : COPY entrypoint.sh /bin/
 ---> ce3d6680bac4
Step 7/8 : RUN chmod 700 /bin/entrypoint.sh
 ---> Running in 10142b8583cb
Removing intermediate container 10142b8583cb
 ---> c27c6cded545
Step 8/8 : ENTRYPOINT [ "/bin/entrypoint.sh" ]
 ---> Running in 3bb0f1944e01
Removing intermediate container 3bb0f1944e01
 ---> fa0a1fe6d348
Successfully built fa0a1fe6d348
Successfully tagged clawplaybook_ansible:latest
WARNING: Image for service ansible was built because it did not already exist. To rebuild this image you must use `docker-compose build` or `docker-compose up --build`.
Creating vm ... 
Creating vm ... done
Creating ansible ... 
Creating ansible ... error

ERROR: for ansible  Cannot start service ansible: cgroups: cannot find cgroup mount destination: unknown

ERROR: for ansible  Cannot start service ansible: cgroups: cannot find cgroup mount destination: unknown
ERROR: Encountered errors while bringing up the project.

[whikloj]

Ok I forgot to run docker-compose build first, so I stopped and removed all running containers.
Then I removed the images.
Then ran docker-compose build

> docker-compose build
Building vm
Step 1/10 : FROM centos:7.6.1810
 ---> f1cb7c7d58b7
Step 2/10 : ENV container docker
 ---> Running in 049a03d4df43
Removing intermediate container 049a03d4df43
 ---> 61e287a5b98d
Step 3/10 : COPY sestatus /bin/
 ---> 64d85322a382
Step 4/10 : RUN chmod 700 /bin/sestatus
 ---> Running in 38df5cc95229
ERROR: Service 'vm' failed to build: cgroups: cannot find cgroup mount destination: unknown

[whikloj]

Does this actually disable SELinux or just the reporting functions?

[birkland]

Since SELinux isn't really running in Docker containers, this just tricks Ansible into passing its "is selinux set up correctly?" test.

[whikloj]

@birkland So I have been able to get this to work, but even though the tomcat8 role shows no errors it seems to fail starting up.

Offending resource: ServletContext resource [/WEB-INF/classes/spring/repository.xml]; nested exception is org.springframework.beans.factory.BeanDefinitionStoreException: IOException parsing XML document from URL [file:/opt/fcrepo/configs/fcrepo-config.xml]; nested exception is java.io.FileNotFoundException: /opt/fcrepo/configs/fcrepo-config.xml (No such file or directory)

I see the files now, and I can restart the service. It seems like at the time the files are not there and the service never restarts. Ideas how to debug this?