Reboot after systemd update?

[Jeff Sadowski]

I think I may need a reboot on a systemd update. Here is why I ask/think that:
systemd runs processes under it thus it is continually running process.
When an update occurs it updates the binary on disk not what is running.
Is that correct?

Is there a way to verify what version is running in memory? So I can compare to last updated systemd and reboot if necessary.

[PabloTwo]

I don't know about determining the running version of systemd in memory, but from recent experience I found it was necessary to do a reboot after upgrading systemd components for other parts of my system to function correctly. Just the other evening I did a "yum update" which contained systemd components.

$ sudo grep "Jul 30" /var/log/yum.log | grep Updated
Jul 30 18:42:53 Updated: systemd-libs-208-21.fc20.x86_64
Jul 30 18:42:58 Updated: systemd-208-21.fc20.x86_64
Jul 30 18:42:58 Updated: libteam-1.11-1.fc20.x86_64
Jul 30 18:42:59 Updated: teamd-1.11-1.fc20.x86_64
Jul 30 18:42:59 Updated: systemd-python3-208-21.fc20.x86_64
Jul 30 18:43:00 Updated: libgudev1-208-21.fc20.x86_64
Jul 30 18:43:00 Updated: systemd-python-208-21.fc20.x86_64
Jul 30 18:43:01 Updated: systemd-journal-gateway-208-21.fc20.x86_64
Jul 30 18:43:01 Updated: libndp-1.3-1.fc20.x86_64

I then, without rebooting, inserted a video DVD to watch, using vlc. vlc couldn't access /dev/sr0, not even as the root user. Rebooted, then vlc had no trouble whatsoever playing the DVD as normal user. There was some other similar incident after a systemd update that caused some nutty behavior that a reboot fixed also, but I can't recall at the moment just what that was.

[Jeff Sadowski]

Not sure PabloTwo if that maybe wasn't earlier updates that might have caused your issue. Sounds like yours might have been able to be fixed by restarting udev that should be possible without a complete system reboot. I can't see any link in vlc not working with a systemd update very peculiar. Most problems can be fixed without a system reboot. I'm not gonna buy into kernel splice yet; I will reboot my system for a new kernel. Some selinux things requite a reboot but my query is if there is a security issue in the current systemd and you upgrade systemd I don't think those fixes take effect till after a reboot unless I am wrong about how systemd works.

I can always check the last time systemd was updated and compare uptime to determine a reboot I guess. I was wanting to write a reboot script that takes place at midnight checking the kernel and systemd to reboot if necessary. I'm not really too worried about a fixed vulnerability being a problem within 24 hours. But I'm not sure I understand systemd enough to know if there is a way to restart it without a reboot.
------------------------------------
Looking further into this I read about restarting systemd
seems others have posted that
"systemctl daemon-reexec"
will restart systemd maybe only a kernel update needs a full reboot.

[stevea]

systemctl --version
yum list installed systemd

[marko]

Is it possible that "systemctl daemon-reload" would restart it? I just tried it and the time stamp on the systemd processes are still old but then I hadn't updated my exe's

[PabloTwo]

I had done "rpm -q --scripts systemd" to see what happens during an install/update.

$ rpm -q --scripts systemd | grep daemon
systemctl daemon-reexec >/dev/null 2>&1 || :
        systemctl daemon-reload > /dev/null 2>&1 || :

Check out the full scripts for yourself, somewhat enlightening. See "man systemctl" for explainations of both commands.

[Jeff Sadowski]

Thank you PabloTwo that looks like it should start the new systemd with that. So maybe a kernel update is all I need to reboot for.

---------- Post added at 08:23 AM ---------- Previous post was at 08:20 AM ----------

Is it possible that "systemctl daemon-reload" would restart it? I just tried it and the time stamp on the systemd processes are still old but then I hadn't updated my exe's

from what I read no "systemctl daemon-reexec" needs done first then "systemctl daemon-reload" so that it loads the new systemd first then restarts all the services.

[renkinjutsu]

Thank you PabloTwo that looks like it should start the new systemd with that. So maybe a kernel update is all I need to reboot for.

---------- Post added at 08:23 AM ---------- Previous post was at 08:20 AM ----------


from what I read no "systemctl daemon-reexec" needs done first then "systemctl daemon-reload" so that it loads the new systemd first then restarts all the services.

systemctl daemon-reload doesn't actually restart anything. It basically keeps systemd running on up-to-date configuration when you edit unit files and the like.

[Jeff Sadowski]

I had learned that no I do not need to reboot with a systemd update. It looks like kernel updates are all that a reboot is required for. Some day even those won't require a reboot. Suse was asking for code to go into 3.19 to allow kernel updates on the fly.

[Dutchy]

[...]
Is there a way to verify what version is running in memory? So I can compare to last updated systemd and reboot if necessary.

With lsof you can see the inodes that a process has opened which you can verify with the ones on disk with ls -i.
If they do not match after an upgrade then it means the old files are still loaded.
lsof also tells you the status of a file by itself so you actually do not need to verify it manually (mem says the file is memory mapped and DEL says the on disk version has been deleted).

Also, for dnf there is a new plugin that will tell you whether you need to restart some program or even reboot after doing updates.
Can't remember its name at the time unfortunately.