----- Original Message ----- > From: "Suvayu Ali" <fatkasuvayu+linux@xxxxxxxxx> > To: users@xxxxxxxxxxxxxxxxxxxxxxx > Sent: Thursday, July 23, 2015 1:07:13 AM > Subject: Re: dnf update vs Software Udpates > > Hi Pete, > > On Wed, Jul 22, 2015 at 05:42:15PM -0500, Pete Travis wrote: > > > > There is a timer unit, `/usr/lib/systemd/system/dnf-makecache.timer`, that > > fires ten minutes after each boot then one hour following the execution of > > each previous run. It triggers > > `/usr/lib/systemd/system/dnf-makecache.service`, a service that executes > > `dnf -v makecache timer`. When that command runs, dnf will check the age > > of the current metadata cache and refresh it if it is older than the value > > of * metadata_timer_sync* (seconds) in `/etc/dnf/dnf.conf`. > > Thank you for this clear and very nice explanation. > > > So, an always-on computer should never have metadata older than 4 hours; in > > practical terms, I think values >2 hours are increasingly unlikely. A > > computer that's been off overnight and turned on in the morning should have > > a fresh cache within 15 minutes of boot. If you have, say, a laptop that > > you power down often and often install or update packages immediately after > > boot, you might adjust the OnBootSec value by copying dnf-makecache.timer > > to /etc/systemd/system/ and editing accordingly. Or, consider appending > > --refresh on an as-needed basis. > > I think this is where things go wrong. OnBootSec handles powerdowns, > what about intermittent connections? In principle, it is quite possible > everytime the timer triggers the makecache service, the connection is > absent. In fact, the probability to hit the sweet spot is directly > determined by the reliability of the connection. So a connection that > is up 50% of the time, will miss 50% of the makecache jobs. Maybe the > makecache jobs can reset the timer to try again in 10 mins in case of no > network connectivity. I think that would make the odds more favourable. > > Essentially I'm suggesting to treat no connectivity as a powercycle. > Hopefully this gives the devs some ideas. > > Cheers, > > -- > Suvayu Can you please file an RFE? Thanks -- Radek Holý Associate Software Engineer Software Management Team Red Hat Czech -- users mailing list users@xxxxxxxxxxxxxxxxxxxxxxx To unsubscribe or change subscription options: https://admin.fedoraproject.org/mailman/listinfo/users Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct Guidelines: http://fedoraproject.org/wiki/Mailing_list_guidelines Have a question? Ask away: http://ask.fedoraproject.org