dnf replacement for yum-cron

[Neal Becker]

drago01 wrote:

> On Mon, Jun 16, 2014 at 5:14 PM, Matthew Miller
> <mattdm at fedoraproject.org> wrote:
>> On Mon, Jun 16, 2014 at 05:06:45PM +0200, drago01 wrote:
>>> > That's not the most descriptiony of all descriptions ever, but if the name
>>> > is any indication, it is just a thing which keeps the cache up to date.
>>> > yum-cron can actually apply updates  [....]
>>> That sounds dangerous ... updates are not really atomic (i.e not at
>>> all) doing them silently in the background is a very bad idea.
>>
>> Yet, it works pretty well most of the time. I've done it at decent scale on
>> production machines with no real issues -- and, most critically, with
>> *fewer* issues than on unpatched systems.
>>
>> Real issues do _occasionally_ occur, but so do bad disks, failed ram, bad
>> offline updates, etc., etc. Fear over lack of atomicity is letting "it's not
>> perfect!" get in the way of real world usefulness.
>>
>> Additionally, these updates aren't _silent_ -- they're logged and there's an
>> e-mailed report.
> 
> Well I meant things like:
> 
> Admin: "OK I will reboot box 'foo'"
> <reboots box 'foo' that was running an update>
> *boom*
> 
> (well actually that case can be "solved" by using systemd-inhibitors
> ... does it do that?)

This server is almost never rebooted, and in many years running yum-cron I've
never had any problem.

What is the difference between automated update vs. a manual update, in terms
of potential for problems anyway?