time jumping back

Michal Jaegermann michal at harddata.com
Sat Feb 22 22:52:00 UTC 2014


Before I will start to write bugzilla entries and the like I wonder if
anybody have seen something like this.  Look at timestamps on a fragment
of a journalctl output:

 Feb 21 09:10:16 some.host dbus-daemon[557]: dbus[557]: [system]
  Activation via systemd failed for unit 'dbus-org.freedesktop.ModemManager1.serv
 ice': Unit dbus-org.freedesktop.ModemManager1.service failed to load: No such fi
 le or directory.
 Feb 21 09:10:16 some.host dbus[557]: [system] Activation via systemd failed for unit 'dbus-org.freedesktop.ModemManager1.service': Unit dbus-org.freedesktop.ModemManager1.service failed to load: No such file or directory.
 Feb 20 23:22:01 some.host systemd[1]: Time has been changed
 -- Subject: Time change
 -- Defined-By: systemd
 -- Support: http://lists.freedesktop.org/mailman/listinfo/systemd-devel
 -- 
 -- The system clock has been changed to REALTIME microseconds after January 1st, 1970.
 
That followed by a number of "systemd[yyy]: Time has been changed"
entries.  A closer examination found that hardware clock was modified
too.

This is a "remote" machine running Fedora 20 with 3.13.3-201.fc20 kernel
and using ntpd for a time synchronization. I failed to notice
immediately what is going on and only wondered why my logs are
terminated at some weird time yesterday.  Only after a while I caught
what is really going on.  Stopping and starting ntpd "jumped" time to
a correct value.

As, so far, this was something I _never_ encountered in this for before
I would be inclined to dismiss that as a hardware/software glitch if not
that detail that from another remote I got email with a definitely weird
and unexpected timestamp.  Checking logs there revealed the following:

 Feb 22 08:01:28 other.host run-parts[18105]: (/etc/cron.hourly) finished mcelog.cron
 Jan 02 23:16:08 other.host systemd[1]: Time has been changed
 -- Subject: Time change
 -- Defined-By: systemd
 -- Support: http://lists.freedesktop.org/mailman/listinfo/systemd-devel
 -- 
 -- The system clock has been changed to REALTIME microseconds after January 1st, 1970.
 Jan 02 23:16:08 other.host systemd[823]: Time has been changed
 -- Subject: Time change
 -- Defined-By: systemd
 -- Support: http://lists.freedesktop.org/mailman/listinfo/systemd-devel
 -- 
 -- The system clock has been changed to REALTIME microseconds after January 1st, 1970.

and so on.  This time it "lost" not some hours but nearly two months.

This other box is running on a quite different, much newer, hardware and
although this is also Fedora 20 with the same kernel it is using chrony
for a time synchronization.  Stopping it and starting corrected that
time but once again I had to use hwclock to bring BIOS from the past.

The same systemd-208-9.fc20 was installed on both machines on January
20th.  There was a big pile of updates on February 18th, and that
included kernel-3.13.3-201.fc20, but it was rather quiet after that.
I have other boxes with a similar software configuration but, so far and
knock-on-wood, I have no other observations like those above.  That is
why I wonder are there any other instances of such time backjumps?
If reporting something in bugzilla then against what?  I have not a
clue.

   Michal


More information about the test mailing list