Hi Pat Riehecky!

 On 2015.07.07 at 16:19:35 -0500, Pat Riehecky wrote next:

> You probably want to alter the startlimitinterval or limitburst[1] in an
> /etc/systemd/system/dhcpd.service
> 
> Odds are an upstream bug is also warranted as this is some unexpected
> behavior.
> 
> Pat
> 
> 
> [1] http://www.freedesktop.org/software/systemd/man/systemd.service.html#StartLimitInterval=
> 

Thanks! That helped.
Actually, I knew that there was that tweak but wanted something better,
because changing that value kind of defeats this "stop service that
respawns too fast" - set it too high and you disable the function, set
it low and after adding more interfaces you might experience it again.
Well, I know this is nitpicking, but I wanted something that won't fail.

Though it solves the problem (just tried it and it does).

But there is a mention of "systemctl reset-failed" on this page, which
*is* the best solution. I just added another dispatcher script which
does that; so now NM both restarts dhcpd and resets failure counter.
The original stop feature works as expected in case something else
happens to dhcpd, and I don't have to worry which is safe value in case
NM will become to aggressive (which might happen in theory on this
system due to vpn starting pppd on connect which might fail again and
again).

I still don't understand why systemctl restart increases failure
counter, must be some misfeature or a bug...


-- 

Vladimir