nispa@hutcs.hut.fi (Tapani Lindgren) (12/04/87)
I have encountered a problem trying to make a shell script that would make incremental backups at nighttime without operator attendance. The problem results from dump(8) program requiring occasional responses from the operator through /dev/tty. The script is run from another script, /usr/adm/daily, under cron control and has no controlling terminal, so it just hangs trying to read /dev/tty. It would be ok if dump just aborted when facing a situation that would require operator intervention. The script should never hang in a loop under any circumstances, because /usr/adm/daily must do other things too and finish after a reasonable time. Currently I have the dump script run a background subshell that sleeps for an hour and then kills the dump script (if it still runs) and all dump processes. This is very complicated, however, and the watchdog process is almost 50% of the whole script. It is also very slow - I would like it to stop immediately if it finds an error, report it to log file, rewind the tape, and let /usr/adm/daily continue its work. Has anyone out there in the Netland have any suggestions of what to do? Can yes(1) somehow be piped to a program that reads /dev/tty? Could dump(8) be modified to abort at errors without any questions? What kind of unattended backup systems do you have? --- Tapani Lindgren, Helsinki Univ. of Technology, CS dept. INTERNET: nispa@hutcs.hut.fi UUCP: mcvax!santra!hutcs!nispa BITNET: nispa%hutcs.UUCP@fingate.BITNET