I have a situation where, to my knowledge, SMF dependencies alone cannot ensure the proper shutdown sequence during a shutdown.
The ADM event manager service need to hang around until after unmounts of the file systems it manages. But unmounts are not controlled by an SMF service. (My service is dependent on file system usr, and makes file system minimal dependent on it. That works on startup only) The path I've been pursuing is... The process catches the signal and does not exit until all the FS's it manages are unmounted. I'm not sure how to deal with this in the SMF script... If my SMF stop method returns 0, SMF apparently knows I lied to it (because the PID still exists?), and cancels the contract, killing my service. If my SMF stop method returns 1, it is put in maintenance mode, and the situation is described as an error in the logs/console, but does work in that the service remains as long as it should. If my SMF stop method tries to wait for the service to exit, that holds up other shutdown processing, so it times out. My questions are: Is there some way for the SMF stop method to basically say "I'm not going to stop, its not an error, don't kill me" Is there some other milestone dependency that would help? Any other ideas/suggestions? (ADM info: http://opensolaris.org/os/project/adm/) Thanks! Joe -- This message posted from opensolaris.org
