Skip to content

Arch Forum 2024-10-24

Participants: Backend devs, Magnus, Zak, Victor

Agenda

  • Retro

Notes

Retro

Retry held using this board: https://www.reetro.app/board/656868c99fa86030049c4286/671a08ded83291d981387af4
Scope/questions:
- What works well/can be improved on the Arch Forum format itself.
- What works well/can be improved on the backend architecture.

The retro concluded with a vote, which items to improve are the most important? As always, lots of good insights/issues. There was no deeper discussion around the issues, nor any time for solutions only to clarify each issue.

These entries got at least 2 votes:

  1. Remove API layer from Moneyflow/Risk, remove separate auth for those services. This is when Majority-internal services call MF/Risk API layer (instead of Service layer like we normally do) (7 votes)
    Update: This is now tracked in https://majority.atlassian.net/browse/MIB-11793
  2. Being able to deploy to stage from branches. The main issue is that devs want to test before PR is merged. (6 votes)
  3. Testing (esp Regr testing) breaks easily and is flaky, making deploys less easy (6 votes)
  4. Lots of PR's show up in majority-backend-pr which makes it hard to see which ones need review. We cannot have the deploy PRs in the same channel as other PRs. (5 votes)
  5. Our way to roll out changes that affect multiple areas doesn't scale very well. This is about features, such as the restricted user feature. Its difficult to foresee how it will affect other areas. (5 votes)
  6. Reverting too complicated (4 votes)
  7. Monthly or Quarterly health checks for services, like DB, memory, logging, monitoring etc. We had something like this before that was good, and could prevent issues such as no-one noticed for a long time that sendtab(?) did not log anything. (3 votes)
  8. Our integration test way is fairly old and almost deprecated. Shakib says there's a more modern way to do it. (3 votes)
  9. Need option to show real raw request/response in stage. To easier know whats going on. (3 votes)
  10. The api layer and authentication - could auth be handled by ingress/gateway? (2 votes)
  11. Too much noise in slack. This is about automated messages. Very easy to miss important ones in the flood of useless messages. (2 votes)
  12. Better way to handle platform updates which breaks renovate. (2 votes)