NodeTextFileCollectorError

warningHosts (node_exporter)

node_exporter is failing to read one or more metric files that a script or cron job drops into its textfile directory.
Severity warning
Source node_exporter 1.x, textfile collector (--collector.textfile.directory)
Key metrics node_textfile_scrape_error, node_textfile_mtime_seconds

What it means

The textfile collector reads every *.prom file in the configured directory on each scrape. If any file cannot be parsed or read, node_textfile_scrape_error is set to 1. The alert fires when that has been true for a sustained period, so a one-off race with a writer does not page anyone.

When parsing fails, the metrics from the broken file are missing. Anything built on them, such as backup freshness, package update counts, or RAID controller status from a vendor script, silently goes stale or disappears.

Common causes

  • The writer is not atomic: node_exporter reads a half-written file.
  • Invalid exposition format: missing newline at the end, bad label quoting, non-numeric values.
  • The same metric name appears in two files with different HELP or TYPE lines, or duplicate series with identical labels.
  • A metric in the file collides with one node_exporter already exports.
  • File permissions or ownership prevent the exporter's user from reading it.

First checks

  1. Find affected hosts:
    node_textfile_scrape_error == 1
  2. The exporter log names the file and the parse error:
    journalctl -u node_exporter --since "1 hour ago" --no-pager | grep -i textfile
    In Kubernetes, kubectl -n monitoring logs <node-exporter-pod> | grep -i textfile.
  3. Validate each file with promtool:
    for f in /var/lib/node_exporter/textfile_collector/*.prom; do
  echo "== $f"; promtool check metrics < "$f"
done
    promtool check metrics also prints lint warnings; focus on the errors.
  4. Check which files are old (a stale writer):
    time() - node_textfile_mtime_seconds > 3600
  5. Check ownership and permissions: ls -l <textfile-dir>.

Fixing it

Fix or remove the broken file, and the error clears on the next scrape. To stop it recurring, make every writer atomic: write to a temporary file in the same directory with a different extension, then mv it into place. The sponge utility from moreutils does this in one step. Give each script its own metric prefix so files never redefine each other's metrics.

Related alerts

← all alert runbooks