• Setting up node exporter
• Setting up Prometheus
• Setting up Grafana
• Custom exporter

Node exporter is a service that collects various system information such as CPU load, ram usage and much more. Fedora already provides the package "node-exporter" and service "node [underscore] exporter":

sudo dnf install node-exporter
sudo systemctl start node_exporter

The service will listen on port 9100.

You can download the latest version of Prometheus from GitHub:

LATEST=$(curl -s https://api.github.com/repos/prometheus/prometheus/releases/latest | jq -cr .tag_name)
wget https://github.com/prometheus/prometheus/releases/download/$LATEST/prometheus-"${LATEST:1}".linux-amd64.tar.gz
tar -xf prometheus-*
cd prometheus-"${LATEST:1}".linux-amd64
sudo cp prometheus /usr/local/bin/

You need to add node exporter to the list of services to scrape, you should edit the /etc/prometheus/prometheus.yml file and add the following Item in the "scrape configs" section:

scrape_configs:
  - job_name: 'linux-server'
    scrape_interval: 5s
    static_configs:
      - targets: ['localhost:9100']

You can start the server by running the command "prometheus". It will listen in port 9090 where you can access the html page. If you are not lazy you can add a systemd service for prometheus (or whatever daemon management you are using). To create a systemd service, add the following content to the file "/etc/systemd/system/prometheus.service":

[Unit]
Description=Prometheus
Documentation=https://github.com/prometheus/prometheus
Wants=network-online.target
After=network-online.target

[Install]
WantedBy=multi-user.target

[Service]
Type=simple
User=prometheus
Group=prometheus
ExecReload=/bin/kill -HUP $MAINPID
ExecStart=/usr/local/bin/prometheus \
  --config.file=/etc/prometheus/prometheus.yml \
  --storage.tsdb.path=/etc/prometheus/data \
  --storage.tsdb.retention.time=30d

Then change permissions, reload the daemons and start the service:

sudo chown prometheus:prometheus -R /etc/prometheus
sudo systemctl daemon-reload
sudo systemctl start prometheus

Prometheus has a web UI accessible by default at port 9090, we will now see how to setup Grafana, a popular and powerful UI to visualize the data.

Install Grafana from your favorite package manager:

sudo dnf install grafana

Fedora already creates a systemd service so you can start it as usual, otherwise you can create the service yourself like we did before.

sudo systemctl start grafana-server

If you are using the default Fedora's SELinux, in order to allow grafana to use the prometheus tcp socket we need to write a custom policy. Create the file "grafana [underscore] websm.te" and add the following content:

module grafana_websm 1.0;

require {
    type grafana_t;
    type websm_port_t;
    class tcp_socket name_connect;
}

# Allow Grafana to connect to ports labeled websm_port_t
allow grafana_t websm_port_t:tcp_socket name_connect;

Then compile It and load It:

checkmodule -M -m -o grafana_websm.mod grafana_websm.te
semodule_package -o grafana_websm.pp -m grafana_websm.mod
sudo semodule -i grafana_websm.pp

Additionally, if you are using a firewall and you want to access the port from another device, you need to enable the port 3000. On firewalld, you can use the following commands:

sudo firewall-cmd --permanent --add-port=30002/tcp
sudo firewall-cmd --reload

You can finally go to "http://localhost:3000" and login with "admin" as the username and password and add prometheus as a source in Home > Connections > Add a new connection and search for prometheus.

You can find already made dashboards for node exporter on the grafana website.

What if you wanted to log something custom to your dashboard, you can easily do this by starting a tcp server and serving the data you want prometheus to read. Then, you need to tell prometheus to fetch your service like we did for node exporter.

One of the simplest way to create a server is using socat. For example, the following script creates a server that serves the system's load average metric for prometheus.

#!/bin/bash

set -e

INTERVAL=5  # seconds
METRICS_FILE="/tmp/power_metrics.prom"
PORT=9200
touch $METRICS_FILE

# Start background HTTP server
socat TCP-LISTEN:$PORT,reuseaddr,fork SYSTEM:"echo HTTP/1.0 200; echo Content-Type\: text/plain; echo; cat \"$METRICS_FILE\"" &

# Metrics writer loop
while true; do
    sleep "$INTERVAL"

    loadavg=$(cat /proc/loadavg | awk '{print $1 }')

    cat <<EOF > "$METRICS_FILE"
    # HELP loadavg Load Average
    # TYPE loadavg gauge
    loadavg $loadavg
    EOF
done

And update the prometheus.yml file:

scrape_configs:
  ...
  - job_name: 'load-average'
    scrape_interval: 5s
    static_configs:
      - targets: ['localhost:9200']

This script runs the tcp server in the background and updates the served content each $INTERVAL seconds and It exposes the loadavg metric with the value of the average system load in the last minute. You can now add a new dashboard in grafana from the dashboard section and clicking "New > New Dashboard > Add Visualization". You select prometheus as a source and you add your custom metric and clock "apply".

You can create the http server in any other way, maybe integrate It with the application you have been working on, and so on.

Travel: Linux Notes, Index