Prometheus metrics¶

DataPress can expose an HTTP endpoint in Prometheus text format for request-volume, latency, and status-code tracking.

Full setup and scrape configuration is documented in Operations › Prometheus metrics. This page covers the [metrics] TOML block.

Build¶

Metrics are opt-in at compile time:

cargo build --release -p datapress-duckdb --features docs,swagger,metrics

When the binary is built without the metrics feature but [metrics] enabled = true is set, the server logs a warning at startup and skips the endpoint.

Configuration¶

[metrics]
enabled = true       # off by default
path    = "/metrics" # scrape path

Key	Default	Notes
`enabled`	`false`	Master switch.
`path`	`"/metrics"`	Scrape path. Served at the bare host root — independent of any configured `server.prefix`. Must start with `/` and not end with `/`.

The endpoint is unauthenticated and sits in front of the [auth] layer, just like the health probes. It exposes only aggregate request counters, never row data. Isolate it at the network layer (private interface, firewall rule, or pod-network-only scraping).

From Python¶

from datap_rs.datapress import DataPress, DataPressConfig, DatasetConfig

config = DataPressConfig(
    backend="duckdb",
    port=8080,
    metrics_enabled=True,
    metrics_path="/metrics",
)

Available metrics¶

Metric	Type	Labels
`datapress_http_requests_total`	counter	`endpoint`, `method`, `status`
`datapress_http_requests_duration_seconds`	histogram	`endpoint`, `method`, `status`

See Operations › Prometheus metrics for a full example, scrape config, and Grafana dashboard notes.