Rainer Heilke wrote: > Darren Reed wrote: > >> Rainer Heilke wrote: >> >>> Should we have a "start" option that has specific implications of >>> transience, and risk the dangers that presents? (This seems to run >>> counter to the original development concepts behind SMF, as >>> discussed here.) >> >> >> >> IMHO, yes. >> At least what I'd like to see is "svcadm start foo" cause foo's >> start method to be invoked - provided that foo is "offline" to start >> with - and update the status to be "online". Yes, it should be >> synchronous (-s) by default so that errors in starting are >> reported "immediately." > > > I mostly agree, with the addition that if start fails, the service > should go offline (with proper logging of the failure, error messages, > whatever), negating the need for a stop or clear. (Just my opinion.)
What I would like to have seen is the ability of a service to specify how to transition from maintenance to offline (which may or may not be the same as "disable") is made, and for the service to be parked in that state until there's an external event (enable/boot.) As it is now, it is assumed that "disable" is the correct way to make that transition, and I'd be tempted to argue that going from "online" to "offline" can be different to a "failed online" to "offline". Although something more would need to be done in order to stop SMF continually looping a service from offline->maint->offline->maint... Darren
