#1112072 ntpsec: ntplogtemp is repeatedly started and stopped rather than let run to periodically log

Package:
ntpsec
Source:
ntpsec
Description:
Network Time Protocol daemon and utility programs
Submitter:
Eric Towers
Date:
2026-02-16 22:19:02 UTC
Severity:
normal
#1112072#5
Date:
2025-08-25 21:08:01 UTC
From:
To:
Dear Maintainer,

I was centralizing logging recently and have been delinting my syslog(s).
I noticed the following frequently repeating pattern

    2025-08-25T20:50:01.874794+00:00 tera003 CRON[908057]: (root) CMD (if [ ! -d /run/systemd/system ] && [ -x /usr/sbin/ntplogtemp ] ; then /usr/sbin/ntplogtemp -o -l /var/log/ntpsec/temps ; fi)
    2025-08-25T20:53:22.138293+00:00 tera003 systemd[1]: Started ntplogtemp.service - Temperature information logger for ntpviz.
    2025-08-25T20:53:23.802596+00:00 tera003 systemd[1]: ntplogtemp.service: Deactivated successfully.

Checking /etc/cron.d/ntpsec-ntpviz , I find
    */5 *    * * * root if [ ! -d /run/systemd/system ] && [ -x /usr/sbin/ntplogtemp
 ] ; then /usr/sbin/ntplogtemp -o -l /var/log/ntpsec/temps ; fi

This is an odd way to use ntplogtemp because the -o option causes the
observed chatty log behaviour every 5 minutes.  However, the -w 300 option
should cause a running instance of ntplogtemp to log every 5 minutes
without as much syslog traffic.

Maybe ntplogtemp should be set up as a system service (and use its internal
timing to set logging frequency) to eliminate this stream of syslog traffic.

#1112072#10
Date:
2026-02-16 22:16:39 UTC
From:
To:
I finally had time to look into this. I agree with it in general, and
have most of the change ready. I posted them here:
https://salsa.debian.org/debian/ntpsec/-/merge_requests/10

This is one of those "devil in the details" situations. I've worked
through a lot of the details, including asking upstream their thoughts
(they're fine with it) and how to convert existing installations.

Remaining things:

- Adding an init.d script for non-systemd users. Someone from Devuan
pointed me to init-d-script(5), which may work. These are non-fork()ing,
and definitely not double-fork()ing/daemonizing, so I need to test to
ensure that works.

- Log rotation. With the timer/cron job approach, the log file is
inherently reopened every time; with a long-running daemon, we need to
restart it (or the daemon needs to support SIGHUP and reopen the files).
I'm not actually sure these are being rotated right now. If not, then
that's its own issue I should try to fix.