Slashdot Mirror
ISS Computer Failure
A number of readers wrote us with news of the computer problems on the International Space Station. Space.com has one of the better writeups on the failure of Russian computers that control the ISS's attitude and some life-support systems. Two out of six computers in a redundant system cannot be rebooted. The space shuttle Atlantis may have its mission extended until the problem is fixed. A NASA spokesman was optimistic that the problem can be resolved; worst-case scenario would be for the shuttle to evacuate everyone onboard the ISS. Engineers are working on the theory (among others) that the failure may have been triggered by new solar panels installed earlier in Atlantis's mission.
Oh, come on. They were Sorny's!
Okay funny guy. :P
:) The reason I bring it up is the hack required for the tivo required a port of monte. Is there any reason that this couldn't become a standard feature? ie, compile the latest kernel fixes, monte in the new kernel, do a service by service restart? I know it's pedantic, but it's then possible to have perpetual uptime, is it not? I"m presuming that the old kernel gets flushed from memory, and since you do a rolling service restart, no one service goes completely down.
That actually brought to mind something really interesting...
We have all of these really cool features in open source software. I do mean REALLY cool too. I run my entire business on FreeBSD servers (no flame wars, just a personal preference!), and amonst all of the really cool things that Linux and FreeBSD can do, I've often wondered something. Why do we have to pull a full reboot for a new kernel?
Don't answer me directly. I know the answer. I know what a kernel is and what it does. Mostly.
I'm making an awful lot of presumptions, and I guess the thought is that if you're going to go to the trouble of doing a rolling service restart, you might as well just cut the power and be done with it...but still. It'd be nice if there's a security fix in the kernel that wouldn't break compatibility with existing running applications to just let you compile, monte to the new kernel, flush the old kernel, and life goes on. Is there a technical limitation that I'm not seeing? Understand I have some FreeBSD-isms going on here with the monolithic kernel vs. kernel+modules. I've run into a nightmare a few times before on a debian box where I've compile a new kernel but forgot to recompile all of the modules, and stuff dies.
Karma: Chameleon (mostly due to the fact that you come and go).
It's the first post. HOW THE FUCK CAN IT BE REDUNDANT?!?