power-hmc-exporter

power-hmc-exporter — Your Entire PowerVM Virtualization Layer, Mapped

Most Power monitoring stops at the operating system: CPU, memory and disks inside each LPAR. The questions that actually cost you time live one layer down, in the virtualization — which shared pool is saturated, which VIOS is the bottleneck for which LPARs, which partition is hammering a physical Fibre Channel port through NPIV, which LPARs sit behind a given SEA and VLAN. Answering them today means hand-correlating lsmap, lshwres, fcstat and the HMC into a spreadsheet.

power-hmc-exporter reads the HMC’s Performance & Capacity Monitoring (PCM) facility and reconstructs that whole topology as live Prometheus metrics — processor and memory pools, the LPAR ↔ VIOS service relationships, and the SEA / NPIV / VSCSI mappings — with the cross-reference labels already resolved. The spreadsheet archaeology becomes a Grafana dashboard.

Processor & memory pools

Stop guessing about pool contention

Per-managed-system shared processor pool utilization and memory pool allocation versus total, entitled-versus-consumed for every LPAR and VIOS, capped-versus-uncapped borrowing against entitlement, and virtual-processor ceilings. See which pool is running tight and which partitions are borrowing from it — before your users feel it, not after.

LPAR ↔ VIOS

Every LPAR’s serving VIOS, resolved for you

Each per-LPAR network and storage series already carries the serving VIOS — the primary SEA VIOS (self-correcting on a SEA failover), the VIOS behind each NPIV physical port, the vhost serving each VSCSI disk. No id→name joins, no lsmap archaeology. A built-in noisy-neighbour view surfaces the VIOS uplinks each critical LPAR actually depends on.

labels already resolved

# the serving VIOS, SEA and VLAN are on the series itself —
# no join against lsmap / lshwres needed
power_hmc_lpar_network_received_bytes{
  system="sys01", lpar="lpar07",
  vlan_id="191", sea="ent10",
  vios="vios02"} 4.81e+08

power_hmc_lpar_vfc_read_bytes{
  lpar="lpar07", fc="fcs1",
  vios="vios02"} 1.66e+09

SEA · NPIV · VSCSI

Network and storage, attributed to the LPAR

SEA throughput broken down by the LPARs and VLANs riding it. NPIV virtual-FC traffic mapped to the physical FC port and the VIOS bridging it, with Top-N consumers per port. VSCSI I/O tied to the serving vhost. The cross-team questions that used to need a war-room — “who is saturating fcs1?”, “which LPAR is flooding this SEA?” — answered on one screen.

Productivity & performance

Hours of correlation become seconds on a dashboard

Drill from a pool, SEA or FC-port spike straight to the offending LPAR and the VIOS bridging it — in seconds, not an afternoon of topas, lsmap and fcstat. Right-size pools and entitlements with evidence, charge back shared-adapter usage per LPAR, and cut the back-and-forth between your AIX, network and storage teams. One central exporter polls the HMC — nothing is installed on your partitions.

Three Grafana dashboards, included

Three ready-to-import dashboards ship in the package — import the JSON, pick your Prometheus datasource, done. From the estate-wide bird’s-eye view, down to a single managed system, down to one partition and the VIOS it leans on.

Specs & Requirements

Deploy in 5 minutes

One central host, no agent on your partitions. Download the tarball, extract, run the installer — it creates the service account, places the binary, installs the systemd unit and an environment template (preserving any existing credentials). Set your HMC connection details and enable the service.

# 1. Extract and install
tar xzf power-hmc-exporter-v0.9.0-linux-amd64.tar.gz
cd power-hmc-exporter-v0.9.0-linux-amd64/
sudo ./install.sh

# 2. Set HMC host / user / password
sudo vi /etc/sysconfig/power-hmc-exporter

# 3. Enable and start the service
sudo systemctl enable --now power-hmc-exporter
journalctl -u power-hmc-exporter -f

Then add the scrape target to prometheus.yml and import the three dashboards under grafana/ via Dashboards → Import.

Licensing

Commercial per-LPAR annual subscription. The price is per monitored LPAR, per year — there is no separate fee for the host the exporter runs on. Volume discounts available for larger estates. A fully-functional 30-day free trial is included; contact legal@aixwatch.com for licensing inquiries.

Frequently Asked Questions