Well, thanks again for pointing me there... It would seem now, that as far as ugly hacks go, the most proper place for my intervention would be patching the /sbin/umountall script, before and/or after the actual umounting calls as appropriate for my two quests. The script itself would be killed after a 5-second timeout, but hopefully a sleeping child process should survive.
There is already a possibility to hang at bootarchive update before that, but it should take within 3600 sec (unless hanging in kernel due to inaccessible pool process, as I had at least once on my box). And it seems that I'd have to revise my UPS-shutdown scripts to make sure they halt to runlevel 0 so that the OS is kept running and the watchdog takes care of the actual reboot when deemed necessary. Halting to 0 is supposed to take the system into firmware, though. So I will have to revise that for SPARCs at least... Perhaps the OS would indeed get unloaded and I'd need some other approach. Though it seems unlikely possible without busting into SVC code and recompiling that. :\ Thanks, share any further ideas please - if you come across any, //Jim -- This message posted from opensolaris.org _______________________________________________ sysadmin-discuss mailing list sysadmin-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/sysadmin-discuss
