Area(s)

area:system

Propose new conventions

Linux Pressure Stall Information (PSI) metrics provide valuable insights into resource pressure on CPU, memory, and I/O subsystems, available on Linux systems with kernel 4.20+.

PSI includes 20 new metric instruments:

• CPU: system.psi.cpu.some.* (avg10, avg60, avg300, total)
• Memory: system.psi.memory.some.* and system.psi.memory.full.* (avg10, avg60, avg300, total for each)
• I/O: system.psi.io.some.* and system.psi.io.full.* (avg10, avg60, avg300, total for each)

Where "some" indicates some tasks are stalled and "full" indicates all tasks are stalled. The avg* metrics represent pressure averages over 10, 60, and 300 second windows, while total metrics track cumulative stall time in microseconds.

PSI metrics can be automatically collected alongside existing host metrics when running on Linux systems with PSI support. On non-Linux systems or when PSI is unavailable, the implementation gracefully degrades with no impact.

Additional Context

• Linux PSI documentation: https://docs.kernel.org/accounting/psi.html
• PSI has been backported to some older LTS kernels
• Prometheus node_exporter already exposes these metrics: https://github.com/prometheus/node_exporter

Open Telemetry integration details

There are issues on this matter in:

• [receiver/hostmetrics] Pressure Stall Information (PSI) from linux hosts opentelemetry-collector-contrib#42779
• [Feature]: Pressure Stall Information (PSI) from linux hosts opentelemetry-go-contrib#8082

And 2 PRs that I am proposing to address these issues:

• [receiver/hostmetrics] Add Pressure Stall Information (PSI) scraper opentelemetry-collector-contrib#43823
• feat: add pressure stall information (PSI) to host instrumentation opentelemetry-go-contrib#8083

Tip

_{React with 👍 to help prioritize this issue. Please use comments to provide useful context, avoiding +1 or me too, to help us triage it. Learn more here.}