Proposed F19 Feature: systemd features

Ales Ledvinka aledvink at redhat.com
Mon Jan 28 15:34:51 UTC 2013


How about crash/inconsistency recovery?

That's either single user mode or multi user networked mode with blocks for anything to finish and nothing new starts and then setting up controlled environment which usually includes no cron interference.

----- Original Message -----
From: "Jóhann B. Guðmundsson" <johannbg at gmail.com>
To: devel at lists.fedoraproject.org
Sent: Monday, January 28, 2013 4:01:40 PM
Subject: Re: Proposed F19 Feature: systemd features



On 01/28/2013 02:56 PM, Marcela Mašláňová wrote: 


On 01/27/2013 04:36 PM, Michael Scherer wrote: 


Le dimanche 27 janvier 2013 à 09:49 -0500, Sam Varshavchik a écrit : 


Jaroslav Reznik writes: 



Announcing various systemd features in one announcement, see bellow: 

= Features/SystemdCalendarTimers = 
https://fedoraproject.org/wiki/Features/SystemdCalendarTimers 

Feature owner(s): Lennart Poettering <lennart at poettering dot net> 

systemd has supported timer units for activating services based on time since 
its inception. However, it only could schedule services based on monotonic 
time events (i.e. "every 5 minutes"). With this feature in place systemd also 
supports calendar time events (i.e. "every monday morning 6:00 am", or "at 
midnight on every 1st, 2nd, 3rd of each month if that's saturday or sunday"). 

So, systemd wants to reinvent cron? 

That's not exactly the same. 

Since a timer can be activated by a unit, or triggered by a inactive 
unit, you could for example run a job only if a unit is running. You can 
also directly express stuff that cron do not do such as running X 
secondes after boot, even if this could be done in cron too ( like 
@reboot, sleep 40 && stuff ). 

I guess also that since systemd support selinux 
( https://fedoraproject.org/wiki/Features/SELinuxSystemdAccessControl ), 
this permit to have a finer grained system for deciding who can or 
cannot disable a timer unit, with a selinux policy. 
On the other hand, cron just permit to edit the whole file, even if I 
guess you can work around this limitation with a clever system 
using / etc/cron.d / . 

I would say that work even before. If I should say according to number of bugs, not many users were using specific SElinux contexts for cronjob tasks. 

No objection to this feature, it might be very powerful for some use-cases. I'm afraid of situation, when half of cronjobs will be converted and half stay as is. Poor admins. 
It only makes sense to migrate the cron jobs that daemon/services bring in with them ( ca 38 out of 99 ) and I cant see how this is supposed to be "poor sysadmins"? It's simple if it's an daemon/service related it's being handled by systemd timers if not it's being handled by cron. The rest of those packages that bring in cron script should be fixed to require cron ( which is not the case now ) 

JBG 

-- 
devel mailing list
devel at lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/devel


More information about the devel mailing list