Page MenuHomePhabricator

Put the PAWS db inside the PAWS cluster?
Closed, ResolvedPublic

Description

Currently PAWS uses a trove db to manage user login data. This has no significant stateful data, and ideally is replaced on a cluster upgrade to prevent the confusing "Respawn" error that results from stale cookies. Lacking data that has any need to persist beyond the life of the cluster, can we simplify the deployment by adding a sql container to it?

Pros:

  • No more trove db to deploy separately
  • Minikube can now more resemble the production cluster, by deploying the same sql container, rather than a sqlite file
  • Cluster more of an atomic unit, preventing discord between a new cluster and an old database (namely the respawn button that does nothing)
  • Collapses the 2 commit deploy in https://wikitech.wikimedia.org/wiki/PAWS/Admin#Deployment to one commit

Cons:

  • Furthered direct experience in the ephemeral nature of all things, are we ready for that?

Event Timeline

rook updated the task description. (Show Details)

https://github.com/helm/helm/issues/2133 suggests that we cannot have substitution in value.yaml, there are yaml anchors but in this case it would be part of a larger string so I don't believe that will work. If anyone more versed in helm than I can find something nicer than the regex a patch or pointer would be welcome. Two variables for one value seems ick.

https://github.com/jupyterhub/zero-to-jupyterhub-k8s/issues/3156 may provide a resolution

rook changed the task status from Open to In Progress.Jul 13 2023, 12:35 PM
rook claimed this task.