Development: Add support for new Docker based test servers with GitHub deployment

[Hialus]

Checklist

General

• I tested all changes and their related features with all corresponding user types on a test server.
• Language: I followed the guidelines for inclusive, diversity-sensitive, and appreciative language.
• I chose a title conforming to the naming conventions for pull requests.

Motivation and Context

This PR builds upon the work done in #5915.
In that PR we reworked the docker compose setup of Artemis to be production ready, so that it could be used for the test servers.
We want to restructure the test servers so they use Docker for everything and to use GitHub instead of Bamboo for deployments to them. This will make maintenance of them easier an allows for more flexibility.
This PR addresses both issues.

Description

For now this PR adds the ability to deploy to the new test servers 4, 5, and 6. They are available under the URLs:

• https://artemis-test4.artemis.cit.tum.de/
• https://artemis-test5.artemis.cit.tum.de/
• https://artemis-test6.artemis.cit.tum.de/

Note the cit.tum.de instead of in.tum.de. The old/legacy test servers will remain available under in.tum.de until the new setup is deemed stable enough.
They all use Atlassian-Prelive and MySQL for now, so they are pretty much identical. In the future they will likely be replaced by different setups, once TS1, TS2, and TS3 are moved to the new system.
The deployment works the same as for the old TS5 via GitHub environments and workflows that have to be approved. Instead of having to reserve a test server on Slack you can now use the lock:artemis-testX labels. They are set by the workflow and ensure that only one PR is deployed at any given time on one test server. Once you are done testing you have to manually remove this label.

List of all changes:

• Labels for deployment (deploy:artemis-testX)
• Add deployment-error label if any error occurs
• For some errors, comments will be added, explaining what went wrong
• Improved the prevention of manually setting test server locks
• Remove lock labels when the PR is closed
• Added docker deploying capability to artemis-server-cli
• Extract deployment into its own file

How the deployment pipeline works:

1. Triggered on labeling of PR
2. Checks if the added label is a deployment label
   • If no: Skip workflow
3. Checks if the docker build is already done for the latest commit
   • If no: Fail workflow and add comment
4. Uses a matrix job to determine which test servers should be deployed to
5. Creates a valid matrix JSON from step 4 output to be used in step 6
6. Deploy to the test servers:
   1. Remove deploy label for TS
   2. Check if TS is already locked
      • If yes: Fail workflow and add comment
   3. Set up SSH keys
   4. Deploy using artemis-server-cli
7. If any error occurred add the deployment-error label

Steps for Testing

Testing this PR is obviously tricky. There are two things you can do:

1. Deploy this PR and go to the test servers and play around with them to find any bugs with them.
   You can also push empty(!) commits to this PR to test edge cases if you want.
2. Test your own PR on the test servers. See below how to do this:

Prerequisites:

• A PR of your own or just use this PR

1. Merge this branch into your branch
2. Wait for the Build & Deploy job to run through
3. Add one or multiple of the deploy:artemis-test4, deploy:artemis-test5, deploy:artemis-test6 labels
4. Wait for the deployment pipeline to run through
5. Test your PR
6. Remove the lock label
7. Revert the merge

Review Progress

Code Review

• Code Review 1
• Code Review 2

Manual Tests

• Test 1
• Test 2
• Test 3
• Test 4

Exam Mode Test

• Test 1
• Test 2

Screenshots

Example error message:

[SolizerCodes]

Tested the Exam Mode with a Test Exam (couldn't get normal exam to work, most likely my fault) and that worked as expected

[SabrinaGlatz]

I tested the Exam Mode on the testing server. It works as expected.

[fschoenberger]

At some points the code could be simplified by using GitHub's Environments feature. Right now, the test environments are hardcoded in the pipeline:

filter-matrix:
    needs: [ get-labels, check-build-status ]
    runs-on: ubuntu-latest
    strategy:
      matrix:
        include:
          # Commented out environments are not yet available and will be enabled in the future

          #- environment: artemis-test2.artemis.cit.tum.de
          #  label-identifier: artemis-test2
          #  url: https://artemis-test2.artemis.cit.tum.de
          #  user: deployment
          #  hosts: artemis-test2.artemis.cit.tum.de
          #  folder: /opt/artemis
          #  host_keys: |
          #    artemis-test2.artemis.cit.tum.de ssh-rsa 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
          #    artemis-test2.artemis.cit.tum.de ecdsa-sha2-nistp256 AAAAE2VjZHNhLXNoYTItbmlzdHAyNTYAAAAIbmlzdHAyNTYAAABBBDRCF0zH3u38zVPNUhJI7zIDsXa3ONiU0GeCv/ybjawkh4alBPnlXtdH0cG1JPtR/Jz/gau92dcqiIFtqdDCDkg=
          #    artemis-test2.artemis.cit.tum.de ssh-ed25519 AAAAC3NzaC1lZDI1NTE5AAAAIMV70ACU6GdLtf1MwUklNltC78UoOPsasZruYh0Ord6n

          # ...

If these values were put into environments this would allow for more flexibility:

• It'd be possible to add and remove test servers without having to change the pipeline
• If someone forks the repository they wouldn't have to change this file (which would always cause conflicts)
• The file would be easier to maintain and more readable.

An example of deploying with labels and environment can be seen here.

It boilds down to using the createDeployment API Call.

[Hialus]

Thanks a lot for your detailed message @fschoenberger!
I'll definitley have a look if this is something we can use for Artemis. Though since this would take quite a while, this would be done in a separate PR.

[Mtze]

Thank you for implementing this 👍 Changes are looking good to me