1458291 - Make Lando more suitable for continuous deployment

Status: NEW

(Conduit :: Lando, enhancement, P3)

Description

[Steven MacLeod [:smacleod]]

Ops is working on a CD pipeline for Lando UI / API in the devsvcdev environment. We should make a few changes to Lando to make it fit better with CD:
 - Consider / Discuss with security combining private / public lando API into a single instance.
 - Automatically migrate the database on startup (if needed)
   - Needs to be rock solid
   - Only single instance does migration (e.g. AUTO_MIGRATOR=1) and other instances check if migrated and sleep / re-check if not (e.g. AUTO_MIGRATOR=0)
   - Add more logging around migrations to ensure they're running correctly.
   - Fail spectacularly if migration fails.
   - Production (Non CD deployments) will have a snapshot before migration, deployment bug needs to indicate migrations present.

Comment 1

[Mark Côté [:mcote]]

Throwing this on the backlog.  Let me know if you think it should fit into the current story list somewhere (maybe when we have an idea of when ops will be done their side).

Comment 2

[Steven MacLeod [:smacleod]]

(In reply to Mark Côté [:mcote] from comment #1)
> Throwing this on the backlog.  Let me know if you think it should fit into
> the current story list somewhere (maybe when we have an idea of when ops
> will be done their side).

I think we should handle this pretty soon (or at least to coincide with ops finishing things on their side). I think targeting somewhere soon after the current productionize story makes sense.

:ckolos, do you have a rough estimate on the timeline here? Are you going to be waiting on these fixes to finish your work to make this happen?

Comment 3

[Chris Kolosiwsky [:ckolos] (ckolos has left the building)]

Ideally, yes, these would all land in order for me to push forward with our side of things. That said, I recently finished an upgrade to the development pipeline that should support CI going forward, but I need some deploys to test (or coordinate deploy testing in a window where it's okay for things to be borked in -dev).

Also, under idea circumstances, I'd like to move -prod to a CI pipeline as well, once we have the process fool-proofed in -dev. That might be wishful thinking, but so be it.