I always thought that SIGKILL is also kind of uncatchable by applications - that's why it is used as a last resort. The Primer actually says that: "The only exceptions are SIGKILL and SIGSTOP, whose default dispositions cannot be changed."
However the logic for an UPS watchdog which would wait for disk unmounts can be inserted into /sbin/umountall script as well, which gets run after killalls. One problem is, when the disks have any sort of problems (i.e. pool is undergoing a self-consistency process and effectively hangs any syscalls such as those issued by "df" or "zfs list") - then testing whether FSes are gone and not hanging in the process might be tricky ;) But these are two different problems, I hope. One watchdog for hanging disks, with a reboot on timer and no checks, and another watchdog for a specific scenario of UPS shutdown. Thanks, //Jim -- This message posted from opensolaris.org _______________________________________________ sysadmin-discuss mailing list sysadmin-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/sysadmin-discuss
