🔧 Unlocking system performance: A practical guide to tuning PCP on Fedora & RHEL

Performance Co-Pilot (PCP) is a robust framework for collecting, monitoring, and analyzing system performance metrics. Available in the repos for Fedora and RHEL, it allows administrators to gather a wide array of data with minimal configuration. This guide walks you through tuning PCP’s pmlogger service to better fit your needs—whether you’re debugging performance issues or running on constrained hardware.

Is the default setup of PCP right for your use case? Often, it’s not. While PCP’s defaults strike a balance between data granularity and overhead, production workloads vary widely. Later in this article, two scenarios will be used to demonstrate some useful configurations.

✅ Prerequisites: Getting PCP up and running

First, install the PCP packages:

$ sudo dnf install pcp pcp-system-tools

Then enable and start the core services:

$ sudo systemctl enable --now pmcd.service
$ sudo systemctl enable --now pmlogger.service

Verify both services are running:

$ systemctl status pmcd pmlogger

🔍 Understanding pmlogger and its configuration

PCP consists of two main components:

• pmcd: collects live performance metrics from various agents.
• pmlogger: archives these metrics over time for analysis.

The behavior of pmlogger is controlled by files in /etc/pcp/pmlogger/control.d/. The most relevant is local, which contains command-line options for how logging should behave.

Sample configuration:

$ cat /etc/pcp/pmlogger/control.d/local

You’ll see a line like:

localhost y y /usr/bin/pmlogger -h localhost ... -t 10s -m note

The -t 10s flag defines the logging interval—every 10 seconds in this case.

🔧 Scenario 1: High-frequency monitoring for deep analysis

Use case: Debugging a transient issue on a production server.
Goal: Change the logging interval from 10 seconds to 1 second.

Edit the file (nano editor used in the examples, please use your editor of choice):

$ sudo nano /etc/pcp/pmlogger/control.d/local

Change -t 10s to -t 1s.

Restart the logger:

$ sudo systemctl restart pmlogger.service

Verify:

$ ps aux | grep '[p]mlogger -h localhost'
$ pminfo -f

Expected output snippet:

records: 10, interval: 0:00:01.000

🪶 Scenario 2: Lightweight monitoring for constrained systems

Use case: Monitoring on a small VM or IoT device.
Goal: Change the logging interval to once every 60 seconds.

Edit the same file:

$ sudo nano /etc/pcp/pmlogger/control.d/local

Change -t 10s to -t 60s.

Restart the logger:

$ sudo systemctl restart pmlogger.service

Confirm:

$ ps aux | grep '[p]mlogger -h localhost'
$ pminfo -f

Expected output:

records: 3, interval: 0:01:00.000

🧹 Managing data retention: logs, size, and cleanup

PCP archives are rotated daily by a cron-like service. Configuration lives in:

$ cat /etc/sysconfig/pmlogger

Default values:

PCP_MAX_LOG_SIZE=100
PCP_MAX_LOG_VERSIONS=14

• PCP_MAX_LOG_SIZE: total archive size (in MB).
• PCP_MAX_LOG_VERSIONS: number of daily logs to keep.

Goal: Keep logs for 30 days.

Edit the file:

$ sudo nano /etc/sysconfig/pmlogger

Change:

PCP_MAX_LOG_VERSIONS=30

No service restart is required. Changes apply during the next cleanup cycle.

🏁 Final thoughts

PCP is a flexible powerhouse. With just a few changes, you can transform it from a general-purpose monitor into a specialized tool tailored to your workload. Whether you need precision diagnostics or long-term resource tracking, tuning pmlogger gives you control and confidence.

So go ahead—open that config file and start customizing your system’s performance story.

Note: This article is dedicated to my wife, Rupali Suraj Patil, who inspires me every day.