[systemd-devel] [EXT] Re: "OnUnitInactiveSec Timer not firing" issue

[Windl, Ulrich]

Hi!

On " Try adding Persistent=yes, it should make it behave as you expect (you
still need to start the unit manually at least once to initially trigger the timer).":

You talk about the service unit, and not about the timer unit, right?

Systemd.timer says (about Persistent): " Note that this setting only has an effect on timers configured with OnCalendar=."
So it shouldn't make a difference if the docs are right.


Regards,
Ulrich


> -----Original Message-----
> From: Andrei Borzenkov <arvidjaar at gmail.com>
> Sent: Monday, July 29, 2024 9:21 AM
> To: Windl, Ulrich <u.windl at ukr.de>
> Cc: systemd-devel at lists.freedesktop.org
> Subject: [EXT] Re: [systemd-devel] "OnUnitInactiveSec Timer not firing" issue
> 
> On Mon, Jul 29, 2024 at 9:33 AM Windl, Ulrich <u.windl at ukr.de> wrote:
> >
> > Hi!
> >
> >
> >
> > I tried to use my first systemd timer, but failed: Either I don’t understand it
> correctly, or there is a bug in systemd (228 of SLES12 SP5):
> >
> > (See also https://unix.stackexchange.com/q/779714/320598)
> >
> >
> >
> > It seems it’s not enough to “enable” the timer, but also “start” it (well, it
> may seem logical from the systemd point of view, but from a cron user’s
> point of view enabling should be enough)
> >
> 
> That is how all systemd units work, it is not specific to timers.
> 
> > Furthermore it seems to be necessary to run the service unit itself,  too
> (assuming it must be enabled also, right?)
> >
> 
> Well, OnUnitInactiveSec is relative to when the unit was last
> deactivated. To deactivate a unit it needs to be active first.
> 
> > But the biggest thing is that systemd seems to lose the point-in-time of the
> last activation, so the timer won’t fire any more (e.g. after package upgrade
> when everything enabled would be re-enabled, and everything started
> would be re-started).
> >
> > But most of all if the system reboots, the timer also won’t fire any more.
> >
> 
> I am not sure about package updates, but if your unit is not started
> on boot by some other means, it will not be started by your timer
> (because the unit will never be deactivated in the first place).
> 
> Try adding Persistent=yes, it should make it behave as you expect (you
> still need to start the unit manually at least once to initially
> trigger the timer).
> 
> Or as described in `man systemd.timer` try adding OnBootSec condition
> to start unit after reboot.
> 
> >
> >
> > So can anybody explain how things should work?
> >
> >
> >
> > My expectation was that an OnUnitInactiveSec timer would fire
> immediately if it never ran, and then every day from that.
> >
> 
> It is not how it is described in the documentation. "Was never active"
> is not the same as "was deactivated" which by definition implies that
> the unit *was* active.