No Reboot Required

technologyreview.com

8 points·soundsop··3 comments

I'm amused to see that someone is trying to sell this. The code has been around for a while, but I don't really see the audience.

Redhat, Novell, etc., aren't going to support this. Not as long as they aren't the ones providing the patches. So larger corporations aren't going to want to purchase this.

If you have a server that absolutely cannot be rebooted, that probably means it's serving a significant amount of traffic 24/7. Of course, most shops like this probably have redundant servers... so rebooting is not, in fact, a problem.

Maybe servers that are taking real-time measurements? And connected to the internet?

I am a Linux sysadmin for a large company and I can say, without any hesitation, that we would love this! In our case, there aren't _that_ many systems that "absolutely cannot be rebooted" but scheduling a reboot on any system does take significant time and effort.

Why don't you think that the Linux vendors won't support this? I can see it taking some time to be introduced, but there'll be a lot of corporate customers out there who would be interested in this and, as long as the process of generating a new patch doesn't take too long, I can't see any reason why this process can't be used by the commercial Linux vendors.

There is just no benefit to this. Even with ksplice you still have to schedule a downtime for each update because any problems due to a bad patch (or, behold, ksplice bug) would cause an unplanned downtime otherwise.

And as grant-parent said, reboots for kernel-updates are a non-issue in larger deployments anyways. If a simple reboot gives you headache then please ask yourself what happens when (not if!) the hardware on that host fails?

Reply on news.ycombinator.com