Halden Metrics ingests time-series data from your fleet, keeps it queryable for years, and pages the right person when a service actually degrades — not when a disk blinks.
Read the docsService statusPrometheus remote-write, OpenTelemetry OTLP, StatsD and a plain line protocol. Agents run as a single static binary with no runtime dependencies.
Raw samples for 15 days, five-minute rollups for two years. Downsampling runs on write, so historical queries stay flat in cost.
Multi-window burn-rate rules with inhibition, so one bad upstream does not produce forty notifications across dependent services.
Point any Prometheus-compatible writer at the ingest endpoint and the series appears in query within a second. No schema registration step.
remote_write:
- url: https://ingest.example-halden.net/api/v1/write
authorization:
credentials_file: /etc/halden/token
Self-hosted deployments ship as a single container with an embedded store. See the documentation for retention tuning and clustering.