chore: 🍰 Automatic Deployment To `stage.ocelot.social` On Push To `master` Branch

Created by: Tirokk

🍰 Pullrequest

Automatic deployment to stage.ocelot.social on push to (merge into) master branch.

See:

• https://docs.digitalocean.com/products/kubernetes/how-to/deploy-using-github-actions/
• https://docs.digitalocean.com/tutorials/build-and-deploy-your-first-image-to-your-first-cluster/
• especially:
  • see example https://github.com/do-community/example-doctl-action
  • see example https://github.com/do-community/example-doctl-action/blob/main/.github/workflows/workflow.yaml

Issues

• fixes #5065 (closed)
• follow up #5093 (closed)

Todo

First Step

• use kubectl or doctl to change the docker image in the deployments

Commands for branded:

kubectl -n default set image deployment/ocelot-webapp container-ocelot-webapp=ocelotsocialnetwork/webapp-branded:latest
kubectl -n default set image deployment/ocelot-backend container-ocelot-backend=ocelotsocialnetwork/backend-branded:latest
kubectl -n default set image deployment/ocelot-maintenance container-ocelot-maintenance=ocelotsocialnetwork/maintenance-branded:latest
kubectl -n default set image deployment/ocelot-neo4j container-ocelot-neo4j=ocelotsocialnetwork/neo4j-community-branded:latest

Commands for not(!) branded:

kubectl -n default set image deployment/ocelot-webapp container-ocelot-webapp=ocelotsocialnetwork/webapp:latest
kubectl -n default set image deployment/ocelot-backend container-ocelot-backend=ocelotsocialnetwork/backend:latest
kubectl -n default set image deployment/ocelot-maintenance container-ocelot-maintenance=ocelotsocialnetwork/maintenance:latest
kubectl -n default set image deployment/ocelot-neo4j container-ocelot-neo4j=ocelotsocialnetwork/neo4j-community:latest

If we really want to use the not changing tag latest then we need kubectl rollout restart deployment/* for all deployments of us in the cluster after setting the image!

Then

• add the change of the Docker image in the deployments of stage.ocelot.social to publish.yaml
  • it looks like we should use explicit versions and avoid latest for production
    • if a pod restarts it may pull a wrong docker image that is to new
      • warnings against using latest:
        • see https://stackoverflow.com/questions/33112789/how-do-i-force-kubernetes-to-re-pull-an-image
• weigh out kubectl set image against kubectl patch
  • I go for kubectl set image
• using helm would be nice either, but the Helm scripts are in our other repository: Ocelot-Social-Deploy-Rebranding
  • impossible yet

General

• change back to publish on master branch push
• add release version, see #5094 (closed)
• set DigitalOcean access token in the repo
• seed db
  • add PRODUCTION_DB_CLEAN_ALLOW and set it correct for staging to true
  • add chrone job to seed db every night?
• add documentation to here or on deploy/rebrand how to change the image name and trigger an image pull on a Kubernetes cluster
• fix ocelotsocialnetwork/webapp:latest and ocelotsocialnetwork/backend:latest on start in cluster, see https://github.com/Ocelot-Social-Community/Ocelot-Social/issues/5065#issuecomment-1179732798
  • !!! do this before this issue in a new issue
    • create new issue #5066 (closed)
• rename maintenance container from maintenance to container-ocelot-maintenance?
  • if not yet create an issue
    • relates https://github.com/Ocelot-Social-Community/Ocelot-Social-Deploy-Rebranding/pull/46
• clean up: search for Wolle