Prometheus-grade observability
for IBM AIX, finally.
The only native Prometheus exporter for AIX on IBM POWER. 22 collectors, 600+ metrics, errpt streamed to Loki — all from a single BFF install. No Linux, no emulation, no workarounds.
The Gap in Your
Monitoring Stack
AIX engineers have always had to work around tools built for Linux. That workaround has a cost.
--------------------------------------------------------------------------- LABEL: DMPCHK_TOOSMALL IDENTIFIER: E87EF18E Date/Time: Fri Apr 10 15:00:00 GMT 2026 Sequence Number: 3049342 Machine Id: ████████████ Node Id: ████████ Class: O Type: PEND WPAR: Global Resource Name: dumpcheck Description The largest dump device is too small. Probable Causes Neither dump device is large enough to accommodate a system dump at this time. Recommended Actions Increase the size of one or both dump devices. Detail Data Largest dump device hd7 Largest dump device size in kb 3145728 Current estimated dump size in kb 9033277 --------------------------------------------------------------------------- LABEL: DMPCHK_TOOSMALL IDENTIFIER: E87EF18E Date/Time: Thu Apr 9 15:00:00 GMT 2026 Sequence Number: 3049341 Machine Id: ████████████ Node Id: ████████ Class: O Type: PEND WPAR: Global Resource Name: dumpcheck Description The largest dump device is too small.
DMPCHK_TOOSMALL error, silently logged to errpt for 4 consecutive days on a production LPAR. No alert fired. Nobody noticed.
No realtime native exporter
Existing workarounds rely on tools like Nmon, or exporters requiring Go, Rust, or specific Python versions — none of which run natively on AIX without painful dependency management. And when they do work, data is rarely real-time.
No centralized error logs
errpt logs are siloed on each LPAR with no way to correlate events across hardware, sequence, or severity. When something breaks, you're manually piecing together context from multiple systems — slowing root cause analysis when it matters most.
No proactive alerts
Without native telemetry, alerting systems can't fire on what they can't measure. Degradation goes undetected for weeks — or months — until users report the symptom. By then, the business impact is already in motion, and you're investigating from a cold trail.
Note from IBM Documentation
IBM's own documentation recommends Prometheus and Node Exporter for Power Systems — but their setup guide targets IBM z16 running RHEL 8.5 only. The pre-compiled binary is for the s390x architecture. It does not run on AIX.
AIXWatch AIX-exporter fills that gap — the only native Prometheus exporter compiled for IBM AIX on POWER hardware. No Linux required, no sidecar, no SNMP bridge. Install via BFF/installp and start scraping metrics in under 5 minutes.
From download to first metric in 5 minutes
Designed by AIX engineers, for AIX engineers. No workarounds, no emulation — just native telemetry that works.
Install
Download the BFF package and run installp on your LPARs. Active and scraping metrics in 2 minutes — no reboots, no kernel modules.
Configure
Point your existing Prometheus instance at your LPARs. Import the included Grafana dashboard — 22 pre-built panels, zero setup time.
Monitor
Build alerting rules on any metric that matters — CPU, memory, disks, errpt events, and hundreds more. Get notified where your team already works — Slack, Microsoft Teams, ServiceNow, PagerDuty, Telegram, email, or any webhook endpoint.
Every metric. Every LPAR.
One dashboard.
Real-time visibility across your entire POWER fleet, in the tools your team already uses.
Blind spots don't stay small
A real-world cascade: how one errpt entry becomes a production incident.
DMPCHK_TOOSMALL logged to errpt
The dump device can no longer accommodate a system crash dump. errpt dutifully records it. No alert fires. No engineer notices — because no tool is watching errpt across LPARs.
LatentSame entry, same time, same silence
The warning repeats itself on schedule. errpt is still siloed on that single LPAR. No correlation across hardware, no pattern detection. The issue grows in the dark through the weekend.
LatentKernel panic on the production LPAR
An unrelated hardware event crashes the system. The kernel tries to write a crash dump to the undersized device — and fails. No dump. No memory snapshot. No root-cause data. The LPAR is down, and you're flying blind into the investigation.
Active incidentYou're paged by users, not by tools
Business apps have been down for hours. Angry tickets escalate to management. Without dump data, your team spends the morning reconstructing the event from partial logs, SAN traces, and guesswork.
Business impactProduction restored — six hours later
SLA breached. The post-mortem reveals errpt had been warning for four consecutive days. The fix would have taken five minutes on Wednesday. It cost six hours of downtime on Monday.
Business impact
A single nmonchart file, opened Monday morning. The 03:00 AM spike is right there — visible, undeniable, and already 24 hours too late.
One errpt entry. Four days of silence. Six hours of downtime. This is what "no native monitoring" actually costs.
No vanity metrics.
Just what the product does.
No inflated stats. No made-up customer count. Just the verifiable facts about what AIXWatch does — and what it doesn't need to do it.
Runtime dependencies
No Python, Java, Go, Rust, or Node.js runtime to install on your AIX LPARs. Pure native C, compiled for IBM POWER — the way AIX software should be.
AIX versions supported
AIX 7.1, 7.2, and 7.3 on IBM POWER 7, 8, 9, and 10 hardware. One BFF package, full native coverage.
OpenMetrics backend
Prometheus, VictoriaMetrics, Grafana Cloud, Mimir, Thanos — or any OpenMetrics-compatible endpoint. No vendor lock-in.
Every metric. Every LPAR.
Five minutes away.
Native Prometheus for AIX, in 5 minutes. No Linux. No emulation. No workarounds.