systemd is systemv times one hundred!
See all
Members (3,315)
Harald Hoyer's profile photo
Michal Sekletár's profile photo
Lennart Poettering's profile photo
systemd's profile photo
Franz-Peter Leitgeb's profile photo
Glenn Holmer's profile photo
John Kwan's profile photo
Clayton Casciato's profile photo
brain power's profile photo
Mauro Toffanin's profile photo
Joshua D'Almeida's profile photo
John Salvatore Fontanelli's profile photo
Guillermo Santos Melgar's profile photo
Tony Nie's profile photo
Johan Heikkilä's profile photo
Anton “tony” Bangratz's profile photo
Max N. Boyarov (zotrix)'s profile photo
Николай Кирчев's profile photo
Darren Wise's profile photo
Raphael Ludwig's profile photo
Tychus's profile photo
Bernd Mueller (IanG)'s profile photo
Dhaval Giani's profile photo
Ivan Ilic's profile photo


Join this community to post or comment

Rui Seabra

Discussion  - 
5 reasons to switch from vSphere to Red Hat Virtualization, plus a 5-minute demo.
It’s time to upgrade or to renew your VMware vSphere licenses. This decision is not easy for a few reasons such as: One or more of your VMware products are EOL (end of life). There are hundreds or…
Add a comment...

Jeff Thompson

Discussion  - 
How do I get started using systemd for an embedded IoT device? I'm going to let 3rd parties develop apps, probably in Java, that the end user can download and run, similar to Android.
Aaron Wright's profile photo
I too used systemd in an embedded environment. I really wish there was a more straightforward way to get to a minimal system, but I just started by disabling everything I could, and then added back just the features I needed.

The really confusing party for me was figuring out what units were actually required and which ones weren't. This is complicated by the fact that numerous units are referenced directly from the code.

It took a lot of trial and error.
Add a comment...
Systemd mount with a condition:

what I would like to do is mount my ntfs data partition during boot process with systemd only if laptop is on AC power. At the moment I am running this script as a startup application in cinnamon, but would like mount during boot.

state=$(acpi a | cut -d' ' -f3 | cut -d -f1)
if [ $state = 'on' ]; then
mount -o umask=0002,gid=1000,uid=1000,nodev,nosuid,nofail,x-gvfs-show /dev/sda5 MNT_LOCATION
Adam D's profile photoAleksandar Bukva's profile photo
So If I do the automount way it will automatically mount when I need it and unmount when it's idle?
Add a comment...

Arnout Engelen

Discussion  - 

i'm trying to make sense of vt switching and systemd-logind, because for me x11 crashes when I switch directly from a wlc-based (wayland) session back to my x11 session.

If I understand correctly, systemd-logind sends a VT_SETMODE ioctl to the kernel to tell the kernel which signals it should use to request freeing the vt resources, and to notify when it can take them back.

So when my wlc-based session registers itself with logind, logind sends the VT_SETMODE ioctl to manage vt switching on behalf of the wlc-based session. It registers SIGRTMIN for the former, and then uses 'sd_event_add_signal(m->event, NULL, SIGRTMIN, manager_vt_switch, m);' to register 'manager_vt_switch' as callback when the kernel requests a vt switch.

However, when I vt switch away from the wlc-based session, I don't actually see manager_vt_switch being called.

Can anybody point out what I'm missing?
Add a comment...

Varun Pandey

Discussion  - 
Hi systemd gurus,

I had some questions regarding sd-journal native APIs.

1. I intend to filter and sniff some journal entries. Are there any performance implication of adding filters to journald?
2. If yes, is there any documentation that talk about how to optimize sd_journal_add_match?
3. Can I use SD_JOURNAL_RUNTIME_ONLY for Storage=persistent?
4. If not, what is the best way to replicate journalctl -f using native APIs? I have tried using sd_journal_seek_tail with sd_journal_wait and sd_journal_next, but I get some stray entries from last boot for some reason.

Thanks in advance for the help!

Zbyszek Jedrzejewski-Szmek's profile photoVarun Pandey's profile photo
+Zbyszek Jedrzejewski-Szmek I agree, let me compile the usecase and report that. Thanks a lot for replying :-)
Add a comment...
Already fixed in Arch ?
Tomasz Torcz's profile photoepsi nurwijayadi's profile photo
Oohh I see .... Thx
Add a comment...

Jared White

Discussion  - 
I am interested in becoming a contributor and adding support for specifying time zones when setting up timers for a service.

Can anyone point me in the direction of who I should talk to about this?
Jared White's profile photoCarey Riley's profile photo
+Jared White how do you do that now? How do you configure times with the present system?
Add a comment...

Björn Bidar

Discussion  - 
I currently use systemd in early boot (initramfs) with systemd-generator-cryptsetup (sd-encrypt on arch here). Is it possible to read the key to encrypt the volume from a tpm module?
Add a comment...

Anthony Kamau

Discussion  - 
Help me understand `BindsTo=`

I have 2 services - call them A and B. I want service B to be started/stopped/restarted when I start/stop/restart service A. To this end, I've used `BindsTo=A.service` in B's service file and this works as intended for start/stop/restart. Good.

Now, I've also got the `Restart=always` in both service files. However, when service B enters `failed` state for whatever reason, it is not restarted - why?

Add a comment...

Wojtek Swiatek

Discussion  - 

When looking at systemd-nspawn I had the problem os passing arguments to "systemctl start systemd-nspawn@mycontainer.service", where mycontainer is a debbootstrap-generated base system.

What is the correct way to pass a parameter (like "--network-bridge=br0") to systemctl?
Elias Probst's profile photoWojtek Swiatek's profile photo
Thanks. I just found where it is mentioned that must be enabled.

I also explicitely enabled systemd-nspawn@machinename.service. It now boots.
Add a comment...

Simon Frei

Discussion  - 
I am having this issue, that I have a user service that is dependent on two local mounts. Should be easily done, so I set
I first tried with Requires as well, but that gives an error, as apparently user services cannot depend on system targets, but no error on After=. Whether this ordering is followed on boot I cannot say, as at the time I login is reached anyway. However on shutdown systemd starts unmounting local mounts and stopping the user script at the same time. As this user script takes its time to exit and accesses disk heavily, unmounting fails with target is busy. I expected that because of the specified After= in the user service systemd would first stop this service, then stop the services of
What did I get wrong there? Any help appreciated.
Add a comment...

About this community

A community for all things +systemd

Rui Seabra

Discussion  - 
Bom sinal! Resta saber em que moldes de aceitação...
 ·  Translate
Projecto de Lei para a resolução do problema do DRM aprovado na generalidade. A equipa do DRM-PT aplaude as bancadas parlamentares do BE, PAN, PCP, PEV e PS que, no passado dia 22 de Dezembro, aprovaram na Generalidade o Projeto de Lei 151/XIII, que “Garante o exercício dos direitos ...
Add a comment...

Rui Seabra

Discussion  - 
Simple experiment with systemd-networkd and systemd-resolved. In my previous post, I wrote about how simple it was to create containers with systemd-nspawn. But what if you wanted to expose to the outside network to a container? The rest of the world can't add mymachines to /etc/nsswitch.conf and ...
Cristian Ciupitu's profile photoRui Seabra's profile photo
Yes, but since systemd-nspawn does not work with selinux it was sadly disabled in the F25 kvm., and inside the containers that is not an issue as they don't have selinux anyway.
Add a comment...

Marti van Lin

Discussion  - 
From trolling to loving systemd ;-)

I was pretty negative about systemd in the past, because I thought it was pretty confusing.

Not any longer. Systemd has come along the way in the meantime.

Now setting up the console font (which was my complaint in the past) has become as easy and logic as it gets.

Just edit /etc/vconsole.conf

Andreas Henriksson's profile photo
fwiw, beware that this part is one of the parts left out in any +Debian based distribution. The systemd-vconsole-setup is very slim, easy and nice. It's however not as ultra-dynamic as the console-setup perl monster which systemd packaging and localectl in Debian (and diverivatives) is integrated with instead. I keep wondering who really needs (or even knows about) the extra flexibility console-setup provides though and if those can't just manually generate the right layout to load.
Add a comment...

Guido Stepken

Discussion  - 
Crashing systemd without problems!?

Can somebody confirm this? Code review necessary?
Guido Stepken's profile photoArc Riley's profile photo
+Guido Stepken I'm referring to the content of the link you shared, not your post specifically.
Add a comment...

Anthony Kamau

Discussion  - 
Slow SSH login on new CentOS 7 LXC Container.

I have read, and suffered the same consequence of restarting d-bus and not restarting systemd-logind, whereby SSH connections take an awfully long time to get to the remote prompt. After much reading, I found this post:

The question then is - "If systemd-logind needs to be restarted if d-bus is restarted, why isn't systemd-logind bound to d-bus?"

For instance, I have a zebra.service file that when restarted/stopped, also restarts/stops ospfd.service and vice versa. In zebra.service, I have a 'Wants=ospfd.service' in the [Unit] section and in ospfd.service I have a 'BindsTo=zebra.service' also in the [Unit] section. Shouldn't this be the default for such interdependent services?

Anyone care to explain why it's not?
David Ford (FirefighterBlu3)'s profile photoMichael Chapman's profile photo
You say "the onus is on upstream caring to fix the issues rather than ignore and close because they haven't replicated it". But that just raises three questions.

First, if they can't replicate the bug, how can you expect them to fix it? Second, if you're hitting the bug so frequently, but few other people are, have you considered that the bug may actually require specific circumstances to hit, and thus be hard to reproduce?. Third, if upstream can replicate the bug, is there any evidence to show they've ignored it?

As I said earlier, I think what you've described is actually a real bug. But I don't think the way you're working around it is particularly good, and, moreover, I see no evidence that you've ever engaged upstream to get it fixed. That's your problem.

G+ is not a support channel.
Add a comment...

Edward Torvalds

Discussion  - 
I am getting 'mount-start-limit-hit' error for my automount ( ). After sometime if the device gets available systemd does not tries to auto-mount it, how can I solve this problem?
device is remote samba share.
Add a comment...

Nate Hardt

Discussion  - 
I have an issue with machined where a vm has crashed, the process is gone, but it still shows up in machinectl output. Is there a mechanism for machined to reconcile this situation?

Things that work: reboot, machinectl terminate vmname.
Things that don't work: restarting libvirtd, systemctl restart systemd-machined.service.

A little background, this is an automated system for vm management. I can always run 'machinectl terminate vm' just before starting a vm as a work around, but it feels like that is a bit of a hack. I'm still working on a clean repro, it happens maybe 1% of the time when killing off the underlying block driver, causing a panic in the VM.

Thanks in advance for any guidance or help!
Nate Hardt's profile photo
Digging around some more, it looks like machined stores its data about what vms are running in /var/run/systemd/machines/. This matches what I am seeing in that my dead VM is still listed.

Looking through the code here: it further seems like on restart it just iterates /var/run/systemd/machines and re-adds everything to its list.

I guess the strange thing is that there is some mechanism to handle this situation as everything works fine most of the time. In looking through the libvirt code that calls TerminateMachine, I found this bug listed as being worked around in the libvirt code:

Listed as fixed but the original reporter still says he can repro it. I guess maybe my simplest option is to just call terminate on a vm before starting it.

I guess I should mention this is on a rhel 7.2 variant.

If this triggers a light bulb for anyone would be happy to hear about it (or even a pointer if there is a better place to ask about this). Thanks!
Add a comment...

Darryl Bond

Discussion  - 
I have a mythtv backend that doubles as a desktop machine. i want to prevent anyone pressing the shutdown button from shutting down or rebooting the box.
The mythtv service runs as the user mythtv.
I have tried
Exec=/bin/systemd-inhibit /bin/mythbackend ...
in the mythbackend.service file. This does nothing and does not show in systemd-inhibit --list. I assume, because it is run as User=mythtv.

I would like to only allow shutdown/reboot if the mythbacked service is stopped.

Colin Guthrie's profile photoDarryl Bond's profile photo
Yep, that worked. My policy now has:
polkit.addRule(function(action, subject) {
if ( == "org.freedesktop.login1.power-off" || == "org.freedesktop.login1.power-off-multiple-sessions" || == "org.freedesktop.login1.power-off-ignore-inhibit" || == "org.freedesktop.login1.reboot-ignore-inhibit" || == "org.freedesktop.login1.reboot-multiple-sessions" || == "org.freedesktop.login1.reboot")
return polkit.Result.NO;
The multiple sessions was the one that fixed it but I included ignore-inhibit.
Add a comment...