Bug#927458: apt-daily.timer Adding random time every 5 seconds

[Boylan, Ross]

What is the result of having time randomization called repeatedly, with our current setup?
If each randomization adds r seconds, r1 the first time, r2 the second, etc, and t is the original time, is the result after repeated (n) randomizations
t+rn (the good case)
t+r1+r2+...+rn (bad case)
If r is always >=, as the message about adding time suggests, the bad case will probably mean the timer never goes off.
If r can be negative, it means that the timer will go off at a wildly random time.

I guess with a naive implementation one figures the time could have been reset by months or years, so you'd completely toss the previous settings if they are set by date and time.  The result would be the good case.

Ross

________________________________________
From: Michael Biebl <biebl at debian.org>
Sent: Tuesday, April 30, 2019 10:44:32 AM
To: Boylan, Ross; 927458 at bugs.debian.org
Subject: Re: Bug#927458: apt-daily.timer Adding random time every 5 seconds

Hi

Am 30.04.19 um 18:22 schrieb Boylan, Ross:
> According to the internet, the time updates every 5 seconds are coming from HyperV, the host hypervisor under which the system in question is running.  Why HyperV acts that way I don't know (guessing: to avoid large time jumps); I agree it seems absurd.  I'm running on a VM under HyperV.
>
> As noted in my original report, these updates can reportedly be suppressed at the hypervisor level but a) it seems that might eventually cause clock drift and b) I have no direct access to the hypervisor (though I could ask someone who does).
>
> I also don't know if HyperV would react badly if the VM time were not in sync with the host (e.g., disable hyperv time updates + install chrony).  I would hope not, but the aggressive update schedule also suggests it might.

I'm not really convinced, systemd is to blame here.
That said, if you feel strongly about this, please consider filing a bug
report upstream at https://protect2.fireeye.com/url?k=e39660c1-bfd60d97-e39647dc-0cc47adb57f0-8b9d12859c154385&u=https://github.com/systemd/systemd/issues
Maybe upstream is willing to add a workaround for this rather
questionable HyperV behaviour.

Regards,
Michael

--
Why is it that all of the instruments seeking intelligent life in the
universe are pointed away from Earth?