DBRepo — A Repository for Databases

tl;dr

docker compose up -d
docker compose logs -f

Build

Build DBRepo from scratch:

make build

Documentation

More documentation can be found online: https://www.ifs.tuwien.ac.at/infrastructures/dbrepo/

Development

CI/CD

We get compute resources in-kind from dataLAB to run our pipeline:

Figure 1. Gitlab runner configuration in the cluster.

Minikube cluster with 6vCPU and 28GB RAM. The CI pipeline is configured as follows in the config.toml:

concurrent = 10
[[runners]]
  executor = "kubernetes"
  environment = [
    "FF_USE_LEGACY_KUBERNETES_EXECUTION_STRATEGY=false"
  ]
  [runners.kubernetes]
    namespace = "{{.Release.Namespace}}"
    privileged = true
    allowed_services = ["docker:24-dind"]
    [[runners.kubernetes.services]]
      name = "docker:24-dind"
      command = [ "--registry-mirror=http://docker-io-mirror:80", "--insecure-registry=docker-io-mirror:80", "--registry-mirror=http://gcr-io-mirror:80", "--insecure-registry=gcr-io-mirror:80", "--registry-mirror=http://ghcr-io-mirror:80", "--insecure-registry=ghcr-io-mirror:80", "--registry-mirror=http://k8s-io-mirror:80", "--insecure-registry=k8s-io-mirror:80" ]
      alias = "docker"
    [[runners.kubernetes.volumes.empty_dir]]
      name = "rundind"
      mount_path = "/var/run/dind"
      medium = "Memory"
    [[runners.kubernetes.volumes.empty_dir]]
      name = "tmp"
      mount_path = "/tmp"
      medium = "Memory"

For each job in the CI/CD pipeline, a pod with three containers is started:

1. build the main build container, you can freely specify any image with image: xyz as base
2. helper the default helper container.
3. svc-0 the Docker-in-Docker sidecar (rootless executed as user rootless/1000) exposing the Docker socket to the build container under `

Note. Only Docker-in-Docker (dind) is allowed as service in the pipeline currently. For each job, a dind-sidecar container svc-0 is started that exposes the Docker socket at /var/run/dind/docker.sock in the build container you can freely configure how you want. We are aware this is not optimal as it exposes root privileges in the cluster.

The full CI/CD pipeline Helm chart is documented in the fda-deployment repository.

Contribute

Contributions are always welcome and encouraged, simply fork the repository and contact Andreas Rauber.

Acknowledgements

We want to thank the following organizations:

• Bundesministerium für Bildung, Wissenschaft und Forschung (BMBWF) for funding during the call "Digitale und soziale Transformation in der Hochschulbildung".
• TU.it for their continuous support in project work, funding and compute resources provided in-kind.

License

This work is licensed under a Creative Commons Attribution 4.0 International License