Sentry’s self‑hosted RAM warning is accurate: the official install is a full observability suite that expects dozens of services and is impractical on a tiny VPS. For teams that only need reliable grouped exception visibility, a much smaller ingest path is viable. Epure implements that path as two containers (a Rust binary handling ingest/API/dashboard and PostgreSQL 16), an Apache‑2.0 license, and explicit refusals: no distributed tracing, session replay, continuous profiling, generic log ingestion, or mobile symbolication; no Redis/Kafka/ClickHouse; and not 100% Sentry protocol parity. The design accepts tradeoffs to avoid install complexity, billing surprises from event metering, heavyweight compatible toolchains, and license friction.
Measured specifics back the claim: on a 2 vCPU / 769 MiB VPS Epure’s idle RSS was roughly 5 MiB and Postgres about 48 MiB (≈53 MiB combined), and under a short burst (≈280-330 req/s) Epure stayed under ~12 MiB while Postgres was the heavier process. Operational knobs include a spike valve that limits raw events per fingerprint (~100/min), a project cap (~5000/hr), and a 2 MB body limit; excess payloads are dropped. Installation uses a pinned Epure image plus postgres:16‑alpine, change the DSN in official Sentry SDKs and disable traces/replay/profiles sampling to match the narrower exception‑only goal. Re‑measure on your host before quoting a minimum.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.