automate production deployment *mostly

• jira: https://ufz-rdm.atlassian.net/browse/TSM-546
• only one release.env file in releases directory
• Add publish-all-things.sh as manual step in https://ufz-rdm.atlassian.net/wiki/spaces/RDM/pages/403243009/Checkliste+Production+Deployment
  • to republish all saved things of configdb
• pipeline check ./compare_dotenv_files.py release.env with .env.example
• deployment/tsm.intranet.ufz.de/deploy.sh
  • remove sudo
  • instead of docker compose pull -q use docker compose create --build to be consistent with prod script
• deployment/tsm.ufz.de/deploy.sh
  • git fetch tag
  • compare tag .env.example with current .env
  • if everything looks good: git checkout tag else?? print Error! do nothing
  • add rm remove-all-data.sh (see deployment/tsm.intranet.ufz.de/deploy.sh)
  • remove RELEASE_ENV_FILE="releases/${SSH_ORIGINAL_COMMAND}.env" git show ${SSH_ORIGINAL_COMMAND}:${RELEASE_ENV_FILE} || exit 1
  • DC="sudo docker compose --env-file .env --env-file ${RELEASE_ENV_FILE}"
    • remove sudo
    • instead use: DC="docker compose --env-file .env --env-file releases/release.env"

Deploy Trigger Script

• Update https://git.ufz.de/rdm-software/timeseries-management/tsm-prod-deployment/-/blob/main/.gitlab-ci.yml?ref_type=heads

Note

• currently we need create --build as we're in the transition to docker volumes
• when docker volumes are completely used, we can switch to docker compose pull -q