On Fri, 14 Nov 2003 20:00:59 EST, Vladimir Parkhaev said: > Hate to stick my nose in ths thread... but how updating SSH daemon > brings down a production system? Well, *that* particular one is unlikely. But I've seen it happen. You install a borked build of ssh (shared lib dependencies are FUN), restart it, your session goes bye-bye, and you can't get back in to fix the runaway sshd that's chewing all the resources.... The more generic point is that in larger shops, you usually need to get *everything* planned and OK'ed in advance, including backout plans. And even then things go wrong. I'm sure I'm not the only sysadmin who's SSH'ed in to an ill box, decided a reboot was needed, and typed 'shutdown -i6 -g0 -y' (runlevel 6 to reboot, zero seconds grace, and don't prompt me), and instead realized 7 seconds later that what the other end *received* was '-i0 -g6 -y' (poweroff with 6 seconds warning), and made a bad situation worse. What *I*'d like to know is how the transposition gremlins know that it's 2AM on a major holiday, or a snowstorm, or other reason that the NOC is running lights-out and nobody's there to push the button to power it back on...
Attachment:
pgp00064.pgp
Description: PGP signature