Usage

Run the exporter with a configuration file:

idrac_exporter --config /etc/prometheus/idrac.yml

It then serves /metrics?target=<bmc> on the configured port (default 9348).

Command-line flags

Flag	Description
--config	Path to the configuration file (default /etc/prometheus/idrac.yml)
--config-watch	Watch the configuration file and hot-reload on change
--verbose	More verbose logging
--debug	Dump raw Redfish JSON responses (implies --verbose)
--trace	Log each Redfish request — method, path, status — without credentials
--once	Collect every configured host once, print the exposition, and exit
--version	Print version and exit

Validating against a BMC

You do not need Prometheus to check that the exporter talks to a BMC correctly.

--once collects every configured host a single time and writes the metrics exposition to stdout, sorted by target so the output is stable and diffable, then exits:

idrac_exporter --config config.yml --once

This is the quickest way to confirm credentials, reachability, and which metric groups a particular BMC exposes.

Tracing requests

When a BMC returns unexpected data, trace what the exporter is doing:

idrac_exporter --config config.yml --once --trace   # request-level: method, path, status
idrac_exporter --config config.yml --once --debug   # full raw JSON of every response

--trace logs every Redfish request (method, path, status) and is token-safe — it never logs credentials or session tokens. --debug additionally dumps the raw JSON body of each response (and implies --verbose), which is useful when a field is parsed as absent and you need to see the exact shape the BMC returned.

Endpoints

Endpoint	Parameters	Description
/metrics	target	Metrics for the specified target. With no target and no default_target, collects all configured hosts (each labeled instance/system); needs honor_labels: true.
/reset	target	Reset internal state for the specified target
/reload		Reload the configuration file
/discover		Prometheus HTTP Service Discovery
/livez		Liveness probe — always HTTP 200, reads no state
/readyz		Readiness probe — always HTTP 200, reads no state
/health		Always HTTP 200 with a JSON body listing the configured BMC hosts
/		Landing page

/reset?target= drops a target's cached collector, forcing fresh Redfish discovery on the next scrape. /reload re-reads the configuration file and resets only the hosts whose credentials changed.

Point Kubernetes and Docker probes at /livez and /readyz. Both are wired to a handler that reads no configuration and no collector state, so a probe can never restart a working exporter because a BMC went away. Never probe /metrics: it collects a BMC per request and can block behind an unreachable one. /health is informational — it is always 200 and its body names each configured host:

{"status":"ok","version":"1.1.2","revision":"b8f6212","hosts":[{"host":"10.0.0.10","scheme":"https"}]}