On Sun, Aug 23 2015, Marc Joliet wrote: > Am Sun, 23 Aug 2015 10:38:23 -0400 > schrieb allan gottlieb <gottl...@nyu.edu>: > >> Thank you marc and fernando (fernando, I think your replies go only to >> marc and not to the group). >> >> So it seems the conclusion is timers can't achieve both >> 1. Run only once a day even if you boot often. >> 2. Not starting for at least 10 minutes after boot >> >> I realize that you can achieve 2 outside the timer by having services >> fired by the timer begin with a 10 minute delay. >> >> However, I thought timers were supposed to achieve 1 & 2, since that is >> what I believe you get with vixie-cron + anacron. >> >> Also, since systemd.cron is based on timers, I would think it would have >> the same problem we are discussing. >> >> allan > > FWIW, this is also mentioned in the anacrontab(5) man page that comes with > systemd-cron: > > "There are subtle differences on how anacron & systemd handle > persistente timers: anacron will run a weekly job at most once a week, with > allways a minimum delay of 6 days between runs; where systemd will try to > run it every monday at 00:00; or as soon the system boot. In the most > extreme case, if a system was only started on sunday; a weekly job will run > this day and the again the next (mon)day. > With carefull manual settings, it would be possible to run the real > anacron binary (not your distro's package) with systemd-cron; if you need an > identical behaviour. > There is no difference for the daily job." > > I have no idea about the last sentence, since I observe the exact same > behaviour with persistent daily timers. > > HTH
Thank you for this. allan
