Context
0020 (Proposed) fixes Sentry for crashes, keeps tracing off until a redaction step exists, and leaves the logging, metrics and health interface open. The backend is now chosen as Encore, which brings tracing, metrics and logs of its own, and the owner picked Better Stack for uptime and logs. Rule 4 of 0020 (tracing off) conflicts with using Encore tracing, so this ADR amends 0020.Decision
- Three tools, one job each. Sentry for crashes and errors (0020 rules 1 to 3 stand); Encore’s built-in tracing, metrics and logs for backend telemetry; Better Stack for uptime monitoring and log aggregation.
- Encore tracing is on for services. Request payloads and URLs are not attached to application spans; the redaction approach is
TODO. Client-side tracing stays off until redaction exists (0020 rule 4 kept for apps). - Vendors stay behind configuration and one reporting function (0020 rule 1).
- Alerts reach the owner’s phone with a runbook link. SLOs come later.
- Sentry and Expo push from day one in mobile.
- No personal data in any telemetry (0071).
Consequences
- The open question in 0020 rule 6 is answered for services by Encore; apps remain open.
- Three vendors to keep configured, one more cost line.