Docker

Image

The published image runs as a non-root user (uid 10001), exposes port 9444, and defaults to --config /etc/pscale_exporter/config.yaml:

docker run --rm \
  -p 9444:9444 \
  -v "$PWD/config.yaml:/etc/pscale_exporter/config.yaml:ro" \
  -e PSCALE1_PASSWORD='your-monitor-password' \
  ghcr.io/fjacquet/pscale_exporter:latest

Mount a writable volume at /var/log/pscale_exporter if you keep file logging (server.logName) enabled; the image pre-creates that directory owned by the runtime user.

Compose stacks

Two compose files are provided:

File	Use
`docker-compose.yml`	Local end-to-end stack — builds the exporter image from source alongside Prometheus, an OTLP collector, and Grafana.
`docker-compose.ghcr.yml`	Pull-based variant using the published GHCR image (no local build).

Bring up the full stack:

PSCALE1_PASSWORD='your-monitor-password' docker compose up -d --build

Service	URL
Exporter	http://localhost:9444/metrics (`/health`)
Prometheus	http://localhost:9090
Grafana	http://localhost:3000 (`admin`/`admin`)
OTLP collector	http://localhost:8889/metrics

Grafana auto-provisions the Prometheus datasource and the PowerScale dashboard from ./grafana/provisioning. Override GF_ADMIN_USER / GF_ADMIN_PASSWORD for anything non-local.

Prometheus scrape & alerts

The bundled prometheus.yml defines two jobs — powerscale (scrapes the exporter directly) and powerscale-otlp (scrapes the collector's re-exposed series). Alert rules live in deploy/prometheus/pscale.rules.yml:

Alert	Condition
`PowerScaleClusterDown`	`powerscale_up == 0` for 5m.
`PowerScaleExporterDown`	`up{job="powerscale"} == 0` for 5m.
`PowerScaleClusterCapacityHigh`	used/total capacity > 85% for 15m.
`PowerScaleQuotaNearHardLimit`	quota usage/hard > 90% for 15m.
`PowerScaleNodeSmartfail`	a node is smartfailing for 5m.
`PowerScaleNodeReadOnly`	a node is read-only for 10m.
`PowerScaleDriveUnhealthy`	any drive in SMARTFAIL/DEAD/STALLED/ERASE/GONE for 5m.
`PowerScaleSyncIQFailed`	a SyncIQ policy's last run failed for 15m (critical).
`PowerScaleActiveCriticalEvents`	unresolved critical OneFS events for 5m (critical).

A minimal standalone scrape config:

scrape_configs:
  - job_name: powerscale
    static_configs:
      - targets: ["pscale-exporter-host:9444"]