Ksplice and kreplace
Posted Nov 27, 2008 13:02 UTC (Thu) by
vonbrand (subscriber, #4458)
Parent article:
Ksplice and kreplace
I worry about any mechanisms that will by force be very rarely used in earnest, and that just can't really be tested for that exact use case before live use. This is just an invitation to trigger Murphy's law.
If the machine is really that indispensable, it should be well protected, and some fail-over provisions should be in place, its applications presumably would be set up to checkpoint and restart; all this regardless of any kernel-replace-while-running wizardry. This problem space has other, well-tested (but sadly much less geeky and exciting) solutions at hand, and has had for some time.
(
Log in to post comments)