Bug#1033548: 02:00 timers dont run in DST transition night

[Marc Haber]

Package: systemd
Version: 252.6-1
Severity: minor

Hi,

aide-common ships the following timer:

[Unit]
Description=Daily AIDE check

[Timer]
OnCalendar=*-*-* 02:00:00
RandomizedDelaySec=2h
Persistent=true

[Install]
WantedBy=timers.target

This didn't run in DST transition night. I think this might be caused by
the clock jumping from 01:59 to 03:00, with 02:00 not existing.

Is there a notation to have a systemd timer run even if the exact time
the timer is supposed to run doesn't happen? I guess this might also be
the case in case of a grossly misticking clock and a timesync daemon
stepping the time, for example, from 01:59:50 to 02:02:00?

Or would be probably be a better idea to trigger a timer if systemd
finds the trigger time in the past without the timer having been
triggered?

Not running at all came as kind of surprise for me.

I might be holding things wrong but I'd like your opinion.

Greetings
Marc


-- Package-specific info:

-- System Information:
Debian Release: 12.0
APT prefers stable-security
APT policy: (500, 'stable-security'), (500, 'unstable'), (500, 'testing'), (500, 'stable')
Architecture: amd64 (x86_64)

Kernel: Linux 6.2.8-zgsrv20080 (SMP w/2 CPU threads; PREEMPT)
Locale: LANG=de_DE.utf8, LC_CTYPE=de_DE.utf8 (charmap=UTF-8), LANGUAGE=en
Shell: /bin/sh linked to /usr/bin/dash
Init: systemd (via /run/systemd/system)

Versions of packages systemd depends on:
ii libacl1 2.3.1-3
ii libaudit1 1:3.0.9-1
ii libblkid1 2.38.1-5+b1
ii libc6 2.36-8
ii libcap2 1:2.66-3
ii libcryptsetup12 2:2.6.1-3
ii libfdisk1 2.38.1-5+b1
ii libgcrypt20 1.10.1-3
ii libkmod2 30+20221128-1
ii liblz4-1 1.9.4-1
ii liblzma5 5.4.1-0.2
ii libmount1 2.38.1-5+b1
ii libp11-kit0 0.24.1-2
ii libseccomp2 2.5.4-1+b3
ii libselinux1 3.4-1+b5
ii libssl3 3.0.8-1
ii libsystemd-shared 252.6-1
ii libsystemd0 252.6-1
ii libzstd1 1.5.4+dfsg2-5
ii mount 2.38.1-5+b1

Versions of packages systemd recommends:
ii dbus [default-dbus-system-bus] 1.14.6-1
ii ntpsec [time-daemon] 1.2.2+dfsg1-1

Versions of packages systemd suggests:
ii libfido2-1 1.12.0-2+b1
pn libqrencode4 <none>
pn libtss2-esys-3.0.2-0 <none>
pn libtss2-mu0 <none>
pn libtss2-rc0 <none>
pn polkitd | policykit-1 <none>
pn systemd-boot <none>
pn systemd-container <none>
pn systemd-homed <none>
ii systemd-resolved 252.6-1
pn systemd-userdbd <none>

Versions of packages systemd is related to:
pn dbus-user-session <none>
pn dracut <none>
ii initramfs-tools 0.142
pn libnss-systemd <none>
ii libpam-systemd 252.6-1
ii udev 252.6-1

-- no debconf information

[Marc Haber]

On Sat, Aug 26, 2023 at 02:41:23PM +0200, Michael Biebl wrote:
> What was/is the output of
> systemctl list-timers
> before and after such a (DST) time jump?

No clue, sorry. I just noticed that the timers didn't run. In aide, I
have worked around that by moving the jo to 01:50. That is a good enough
fix for me for the moment.

> I.e., is there a way to reproduce this somehow?

I think it would be possible by establishing a timer to 02:00 and
setting the time to DST night.

Greetings
Marc


--
-----------------------------------------------------------------------------
Marc Haber | "I don't trust Computers. They | Mailadresse im Header
Leimen, Germany | lose things." Winona Ryder | Fon: *49 6224 1600402
Nordisch by Nature | How to make an American Quilt | Fax: *49 6224 1600421

[Marc Haber]

On Sun, Oct 29, 2023 at 11:24:54PM +0100, Michael Biebl wrote:
> I think the above upstream issue pretty much describes what you are seeing,
> so marking as forwarded accordingly.

It does. It also nicely reflects upstream's attitude "Yes, you're holding
it wrong, and we don't care how ugly your workarounds might be as long
as we don't have to code anything".

Greetings
Marc, who will probably just avoid 1.59-3.01 am for the rest of life