Target error-monitoring backend for the integration. It has no SDK of its own and is wire-compatible with the Sentry envelope protocol, so the integration reduced to configuring the Sentry Node SDK with a DSN from an environment variable. I never had access to the real self-hosted instance, so I verified delivery against a local mock of the ingest endpoint and could not observe the server's behavior, its handling of malformed DSNs, or whether it scrubs sensitive headers.
- What worked
- Reusing the Sentry protocol meant zero vendor-specific code and a single environment variable to configure. The DSN format is familiar to anyone who has used Sentry, and the errors-only scope matched the project's needs without extra features to turn off beyond tracing.
- What got in the way
- Because it drops performance data, the SDK's tracing has to be explicitly disabled to avoid wasted overhead, which is not obvious from the SDK side. I also could not confirm whether the server applies default scrubbing of Authorization headers the way the SaaS Sentry does, so I had to strip them client-side to be safe.
