Skip to content

perf(roles/aide, roles/librenms, roles/system_update): skip the check right after an AIDE re-baseline - #750

Merged
markuslf merged 1 commit into
mainfrom
perf/aide-skip-check-after-rebaseline
Nov 16, 2026
Merged

markuslf merged 1 commit into
mainfrom
perf/aide-skip-check-after-rebaseline

Conversation

@markuslf

Copy link
Copy Markdown
Member

After a successful AIDE re-baseline, update-and-reboot, security-update, librenms-daily, the dpkg hook and update-db.yml of the aide role no longer start aidecheck.service. aide --update already scans and compares the whole file system before it writes the new database, so the check right after it only repeated that scan and found nothing. Now they just run systemctl reset-failed aidecheck.service. After a finding or a failed update, the check still runs as before.

The update is still scanned and accepted right away (aide --update + mv), and the window between the package transaction and that scan stays the same. On a 1-vCPU Rocky 8 host, the extra check was what set off the PSI I/O warning after the nightly update.

Tests:

  • molecule test --scenario-name aide on Rocky 8, 9, 10 and Debian 12, 13: all stages green. verify.yml now also asserts that a re-baseline (handler, unattended-upgrades) starts no check, and that a check afterwards is clean.
  • End to end with update-and-reboot on the same five platforms (aide 0.16 to 0.19.2), with real package and kernel updates and a stale failed aidecheck.service: no second check, Result=success, aide-logfile reports the accepted update as OK, and a check by hand afterwards is clean.
  • librenms-daily on Rocky 9: no check after a re-baseline. With drift before daily.sh, it still starts a check, which fails.

… right after an AIDE re-baseline

`aide --update` compares the whole file system like a check before it writes the new
database, so the check that followed it scanned everything a second time only to find no
differences, and on a small host it was the main cause of I/O pressure alerts after the
nightly update. After a successful re-baseline the scripts and the role now only reset a
failed state of aidecheck.service. After a finding or a failed update the check still runs.

Verified with molecule (aide) on Rocky 8, 9, 10 and Debian 12, 13, and end to end with
update-and-reboot on the same platforms and librenms-daily on Rocky 9: no second check,
aide-logfile reports the accepted update as OK, a check afterwards is clean.
@markuslf
markuslf merged commit ee7b41f into main Nov 16, 2026
11 checks passed
@markuslf
markuslf deleted the perf/aide-skip-check-after-rebaseline branch November 16, 2026 06:48
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant