abh.textfile_collectors

Ansible collection that installs textfile collector scripts for Prometheus node_exporter, each with a systemd timer writing a .prom file into the textfile collector directory.

Scripts under roles/*/files/upstream/ are unmodified copies from node-exporter-textfile-collector-scripts (Apache-2.0, license in roles/*/files/upstream/LICENSE); the others are local to this collection. The distro package prometheus-node-exporter-collectors would be simpler, but on Ubuntu 24.04 it depends on the distro prometheus-node-exporter, which conflicts with a node_exporter installed by prometheus.prometheus.

Roles

apt_info

Runs apt_info.py every 15 minutes, after dpkg runs, and after a successful apt update. It writes apt.prom with:

• apt_upgrades_pending, apt_upgrades_held (by origin and arch)
• apt_autoremove_pending, apt_packages_obsolete
• apt_packages_per_origin
• apt_package_cache_timestamp_seconds
• node_reboot_required (whether /run/reboot-required exists)

The script doesn't run apt update; Ubuntu's apt-daily.timer does.

Variables (see roles/apt_info/defaults/main.yml):

Variable	Default
apt_info_textfile_dir	{{ node_exporter_textfile_dir }}
apt_info_interval	15min
apt_info_exclude	[]
apt_info_install_dir	/usr/local/lib/textfile-collectors

With prometheus.prometheus.node_exporter in the same play, the directory comes from that role. Elsewhere, set apt_info_textfile_dir.

- hosts: all
  become: true
  roles:
    - prometheus.prometheus.node_exporter
    - abh.textfile_collectors.apt_info

conntrack_stats

Runs conntrack_stats.py every minute. It writes conntrack.prom with node_nf_conntrack_entries, node_nf_conntrack_entries_limit and the node_nf_conntrack_stat_* counters (drop, early_drop, insert_failed, ...), summed over all CPUs, under the same names and types node_exporter uses.

node_exporter reads those counters from /proc/net/stat/nf_conntrack, which only exists on kernels built with CONFIG_NF_CONNTRACK_PROCFS. Ubuntu's kernels (24.04, 26.04) are not, so node_exporter's conntrack collector fails there; this role gets the counters from conntrack -S over netlink instead. Use it only on such hosts, and disable node_exporter's conntrack collector there, or the entries metrics are collected twice and the textfile collector reports an error.

Variable	Default
conntrack_stats_textfile_dir	{{ node_exporter_textfile_dir }}
conntrack_stats_interval	1min
conntrack_stats_install_dir	/usr/local/lib/textfile-collectors

write_textfile

Installs write-textfile, the helper the other roles' units use to publish a script's output as a .prom file. The textfile directory usually belongs to the exporter user while the units run as root, so it creates the temporary file with O_EXCL|O_NOFOLLOW and renames it into place. The other roles depend on it; there's no need to list it in a play.

Installing

From git, in collections/requirements.yml:

collections:
  - name: https://gitea.develooper.com/ask/ansible-textfile-collectors.git
    type: git
    version: v0.2.0

Updating the upstream scripts

UPSTREAM records the upstream commit and the sha256 of each copied file.

make update            # pin the upstream default branch
make update REF=<sha>  # or a specific commit, branch or tag
make check             # local files match UPSTREAM; collection builds
make verify            # the pinned commit still serves those files

make update refuses to run if a local copy was edited, prints the upstream commits that touched the files, and leaves the change for review. Bump version in galaxy.yml and tag a release after updating.