open-feature
[CHORE] Identify and fix single points of failure in flagd architecture
The only single point of failure allowed should be unavailability of flagd Sync service during initialization.
I think if anything, this feels like something we could just mention in https://flagd.dev/architecture/, since there's so many ways to deploy flagd, it could just be something we mention in diagrams (consider the different points of failure in a given architecture or topology).
Was there something else you were thinking here?
The intent here was that after this issues is closed, every flagd provider has only one single point of failure - unavailability of flagd Sync service during initialization. All other potential problems should be recoverable if conditions change. So the owner of the issues would be responsible for:
- auditing current points of failures across all the providers and filing relevant issues
- driving or delegating creation of ADRs to address unwanted points of failure