On 11/21/2019 05:53 AM, Christian Hesse wrote:

I've created systemd configuration overlay snippets for this, for example /etc/systemd/system/man-db.timer.d/RandomizedDelaySec.conf:

[Timer] RandomizedDelaySec=30min

Create a file for every timer you want to delay.

Thank you Ralph & Christian, I'll do that. Something to just keep them all from firing when I boot would be nice. I wonder why systemd doesn't do this by default? As more an more things have been turned over to systemd timer, it should have become evident that they were all going to start stacking up on boot.... Bad Juju... Logrotate and man-db can be run at 4:00 am and be fine. If I update a slew of man-pages, I'll call it directly. I see what I can do with RandomizedDelaySec=. Thanks again. -- David C. Rankin, J.D.,P.E.

On Thu, 21 Nov 2019 at 21:02, NTS wrote:

On Thu, 21 Nov 2019 at 18:13, David C. Rankin wrote:

...

On 11/21/2019 05:53 AM, Christian Hesse wrote:

...

I've created systemd configuration overlay snippets for this, for

example

...

...

/etc/systemd/system/man-db.timer.d/RandomizedDelaySec.conf:

[Timer] RandomizedDelaySec=30min

Create a file for every timer you want to delay.

Thank you Ralph & Christian,

I'll do that. Something to just keep them all from firing when I boot would be nice.

Can't you make these services depend on another one which you write to start a certain number of minutes after boot?

Regards, NTS

You could, but if you're going to modify the services anyway, then a randomized delay seems betters. That way, they are not only delayed, they are also spread out. -- Maarten

"David C. Rankin" on Thu, 2019/11/21 12:13:

I wonder why systemd doesn't do this by default?

It's not systemd to blame. The timer unit files are shipped by the respective projects, like util-linux, man-db, mlocate, shadow, logrotate, ... -- main(a){char*c=/* Schoene Gruesse */"B?IJj;MEH" "CX:;",b;for(a/* Best regards my address: */=0;b=c[a++];) putchar(b-1/(/* Chris cc -ox -xc - && ./x */b/42*2-3)*42);}

Hi Chris,

David C. Rankin wrote:

...

I wonder why systemd doesn't do this by default?

It's not systemd to blame. The timer unit files are shipped by the respective projects, like util-linux, man-db, mlocate, shadow, logrotate, ...

systemd.timer(5) says AccuracySec= ... Within this time window, the expiry time will be placed at a host-specific, randomized, but stable position that is synchronized between all local timer units. This is done in order to optimize power consumption to suppress unnecessary CPU wake-ups. So I think it is systemd's `fault' that they all collide. I think its theory is flawed. The CPU here will happily move from doing nothing to doing slightly less for a unit and back again without getting hot and altering the fan speed. But ig the typical timer units collide then the fan speed moves up through the gears and keeps going at more than idle speed for some time, long after the units have finished and the CPU clocked back down. That must impact power consumption too. -- Cheers, Ralph.