node_exporter¶
The Prometheus exporter for Linux servers: offers CPU, memory, disk, file system and network values on port 9100. On Debian and Ubuntu the program is called
prometheus-node-exporter.

Syntax¶
node_exporter [--web.listen-address=:9100] [--collector.NAME | --no-collector.NAME]
Options¶
| Option | Description |
|---|---|
--web.listen-address= |
Address and port, default :9100 |
--collector.textfile.directory= |
Also read *.prom files from this directory |
--collector.systemd |
Enable the collector for systemd units (off by default) |
--no-collector.NAME |
Switch off a collector, e.g. --no-collector.wifi |
--collector.disable-defaults |
Start with no collectors, then enable single ones |
--web.config.file= |
File with TLS and basic auth settings |
On Debian and Ubuntu the options go into ARGS="…" in /etc/default/prometheus-node-exporter; then sudo systemctl restart prometheus-node-exporter.
Examples¶
A few values straight from the exporter:
curl -s localhost:9100/metrics | grep -E '^node_(load1|memory_MemAvailable_bytes|memory_MemTotal_bytes) '
node_load1 0.18
node_memory_MemAvailable_bytes 2.61210112e+09
node_memory_MemTotal_bytes 4.11205632e+09
Is the service running:
systemctl status prometheus-node-exporter --no-pager | head -3
● prometheus-node-exporter.service - Prometheus exporter for machine metrics
Loaded: loaded (/usr/lib/systemd/system/prometheus-node-exporter.service; enabled; preset: enabled)
Active: active (running) since Mon 2026-10-26 10:42:17 CET; 2h 11min ago
Free space on the root file system:
curl -s localhost:9100/metrics | grep 'node_filesystem_avail_bytes.*mountpoint="/"'
node_filesystem_avail_bytes{device="/dev/sda2",device_error="",fstype="ext4",mountpoint="/"} 2.4823209984e+10
Values are always in base units: bytes, seconds. 2.48e+10 bytes are about 23 GiB.
An own value via the textfile collector:
echo "backup_last_success_timestamp_seconds $(date +%s)" > /tmp/backup.prom
sudo mv /tmp/backup.prom /var/lib/prometheus/node-exporter/
curl -s localhost:9100/metrics | grep ^backup_
backup_last_success_timestamp_seconds 1.793007737e+09
An alert rule like time() - backup_last_success_timestamp_seconds > 86400 then reports a backup that hasn't run for a day.
Collectors that found nothing:
curl -s localhost:9100/metrics | grep '^node_scrape_collector_success.* 0$'
node_scrape_collector_success{collector="bonding"} 0
node_scrape_collector_success{collector="fibrechannel"} 0
node_scrape_collector_success{collector="infiniband"} 0
node_scrape_collector_success{collector="tapestats"} 0
node_scrape_collector_success{collector="zfs"} 0
0 isn't an error here: the server simply has no network bonding, no Fibre Channel, no InfiniBand, no tape drive and no ZFS.
When do I need it?¶
- On every server that Prometheus should watch
- To check by hand which values a server delivers, before writing a rule
- To feed your own values into Prometheus, such as the time of the last successful backup
Own metrics with the textfile collector
A script writes lines like backup_last_success_timestamp_seconds 1793007737 into a
.prom file in the textfile directory. The node exporter delivers them with every scrape:
ideal for cron jobs. Write to a temporary file first and rename it, so the exporter never
reads half a file.
Port 9100 is open to everyone
The node exporter has no login by default. Allow the port only from the Prometheus server,
for example with ufw: sudo ufw allow from 192.0.2.10 to any port 9100 proto tcp.
Distribution differences¶
| Debian | Ubuntu | RHEL | Arch | |
|---|---|---|---|---|
| Package | prometheus-node-exporter |
prometheus-node-exporter |
not in the repositories: release archive or container prom/node-exporter |
prometheus-node-exporter |
| Program and service | prometheus-node-exporter |
prometheus-node-exporter |
node_exporter |
prometheus-node-exporter |
| Options in | /etc/default/prometheus-node-exporter |
/etc/default/prometheus-node-exporter |
own systemd unit | /etc/conf.d/prometheus-node-exporter |
Related commands¶
prometheus · promtool · curl · systemctl