As of today each lister has its own db. This is mostly irrelevant as they use one table each (except for lister-debian).
This however is a pain in the neck to deploy.
In regards to db creation for example.
- Stop lister/loader-debian services
- Change this to merge all dbs into one (swh-lister)
- Grant guest user read-only access
- Unify the rights to the one user swh-lister (with write permission)
- [ ] Rename swh-lister to swhlister (and figure out why connection refuses to work with swhlister as user)
- Keep swh-lister for now
- Adapt puppet manifests accordingly
- Deploy changes on production and start listers
- worker01 -> happy
- other workers
- Start back loader-debian service
- Drop old dbs (lister-github, lister-pypi, lister-npm, lister-gitlab, lister-bitbucket)
- Update the db inventory doc