dell-bios-profile-exporter

A Prometheus-compatible exporter and Helm chart that monitor the BIOS System Profile attribute on Dell PowerEdge servers via the host-local racadm utility (through the iDRAC Service Module), and alert when it drifts from a target value (default PerfOptimized).

What it does

A DaemonSet runs a small Go exporter on each selected node. On every poll the exporter calls racadm inside the host PID and mount namespaces via nsenter, reads the BIOS.SysProfileSettings.SysProfile attribute, caches it (poll interval default 60s), and serves it as Prometheus metrics on port 9101. A small amount of static inventory (Service Tag, model, iDRAC version) is refreshed on a slower interval and attached as labels.

An alert fires when the current profile drifts from the configured target (default PerfOptimized), when the exporter cannot read the value, or when the last successful read is too old. Because everything goes through the host-local iDRAC Service Module, the exporter needs no network access to iDRAC and no credentials: it talks to the local management interface the same way an on-host administrator would.

Architecture

DaemonSet pod --nsenter--> host racadm --iSM (KCS/USB-NIC)--> iDRAC --> BIOS

Why this way: querying iDRAC over the network with Redfish would require iDRAC credentials stored in the cluster and network reach from worker nodes into the management plane, both of which are commonly disallowed and were deliberately avoided here. The syscfg tool (from Dell's deployment toolkit) is not generally packaged for Ubuntu, so it is not a portable option. Host-local racadm talking to the iDRAC Service Module uses the in-band channel (KCS or the internal USB NIC) that the iDRAC Service Module already maintains on the host, so it needs no network path to the BMC and no secrets at all. The exporter simply enters the host namespaces and runs the same racadm the node administrator would run.

Requirements

• Kubernetes 1.24+.
• Helm 3.x or 4.x.
• On each target node:
  • A Dell PowerEdge server, 12th generation (12G) or newer.
  • The iDRAC Service Module installed (dcism), with its daemon dcismeng running and connected to the iDRAC.
  • The racadm binary available on the host (default path /opt/dell/srvadmin/sbin/racadm).
• One monitoring stack already running in the cluster, either kube-prometheus-stack (Prometheus Operator) or k8s-victoria-metrics-stack (VictoriaMetrics Operator).

Contents

• exporter/ - the Go exporter (DaemonSet container).
• chart/ - the Helm chart (DaemonSet, monitoring CRDs, alerts, dashboard).
• examples/ - ready-made values for common setups.
• scripts/verify.sh - runs lint/template/unit/kubeconform checks.

Build the image

cd exporter
docker build --platform linux/amd64 --build-arg VERSION=0.1.5 \
  -t ghcr.io/cicdteam/dell-bios-profile-exporter:0.1.5 .

Test the chart without installing

The chart works with both Helm 3.x and Helm 4.x.

helm lint chart/
helm template chart/
helm unittest chart/
./scripts/verify.sh

Note: under Helm 4 the helm-unittest plugin installs with helm plugin install https://github.com/helm-unittest/helm-unittest --verify=false.

Install in an air-gapped environment

# On a connected host, pull the published chart from the OCI registry:
helm pull oci://ghcr.io/cicdteam/charts/dell-bios-profile-exporter --version 0.1.5
# copy dell-bios-profile-exporter-0.1.5.tgz into the closed network, then:
helm install dell-bios ./dell-bios-profile-exporter-0.1.5.tgz -f my-values.yaml

The container image must be mirrored into the private registry separately (for example with docker save / skopeo copy into your internal registry).

Metrics

Metric	Type	Labels	Meaning
dell_bios_sys_profile_info	gauge (=1)	node, profile, service_tag, model, idrac_version	Info metric; the current profile value is carried in the profile label.
dell_bios_sys_profile_matches_target	gauge	node, target	1 if the current profile equals the target, else 0.
dell_bios_racadm_success	gauge	node	1 if the last poll succeeded, 0 if it failed.
dell_bios_racadm_duration_seconds	gauge	node	Duration of the last racadm call, in seconds.
dell_bios_racadm_errors_total	counter	node, reason (timeout, exit_code, parse_error, nsenter_failed)	Count of failed racadm calls by reason.
dell_bios_last_scrape_timestamp_seconds	gauge	node	Unix time of the last successful poll.
dell_bios_exporter_build_info	gauge (=1)	version, go_version	Build information for the running exporter.

Grafana dashboard

The chart ships a Grafana dashboard (chart/dashboards/dell-bios-profile.json), auto-discovered by the Grafana sidecar when dashboard.enabled=true. It shows node counts (monitored / on-target / drifting / exporter-failing), profile distribution, per-cluster breakdowns and a per-node detail table. Screenshot below uses synthetic data:

For node prerequisites (installing iSM and racadm), chart installation, configuration, alerts, the Grafana dashboard and troubleshooting, see chart/README.md. Russian: README.rus.md.