W dniu 20 marca 2011 20:54 użytkownik cornel panceac <cpanceac@xxxxxxxxx> napisał: > > > 2011/3/20 Genes MailLists <lists@xxxxxxxxxxxx> >> >> On 03/20/2011 03:43 PM, Michał Piotrowski wrote: >> >> > >> > In a stable distribution such situation would not be possible because >> > of the whole testing procedures etc. >> > >> > But of course there should be a way to avoid such situation. I hope >> > that Lennart will find a good way to solve this problem soon. >> > >> > >> > >> >> Then make the variable timeout default to 30 secs - so worst case >> services get started and system runs ... individual services which >> actually require the pre-requisite - can change the timeout variable to >> demand pre-req is satisfied - of course even these will presumably fail >> (hopeflly gracefully). >> >> So the answer seems to be implement a timeout for pre-requisites and >> default it to something sensible - ... fixup any upstart files as needed >> - since the systemd unit files are all new anyway - we should be fine >> with this. >> > although i understand the general idea about systemd, i believe that in > fedora's case there has to be a plan which says who starts when, and what > requires what. then, if thngs don't go that well, as it happened with this > rsyslog update, the plan can be imporved. > -- > test mailing list > test@xxxxxxxxxxxxxxxxxxxxxxx > To unsubscribe: > https://admin.fedoraproject.org/mailman/listinfo/test > Unfortunately it is not so simple. For example we have a service After=syslog.target network.target and for some reason network configuration takes longer than 30 seconds so the service does not start when default timeout will be short and will cover all dependences. -- Best regards, Michal http://eventhorizon.pl/ -- test mailing list test@xxxxxxxxxxxxxxxxxxxxxxx To unsubscribe: https://admin.fedoraproject.org/mailman/listinfo/test
