Maybe we should explicitly schedule rebooting our fleet every so often

utcc.utoronto.ca

19 pointsspeckx6 comments

“If it hurts, do it more often.”

* does not apply to shooting pains

thanks for letting everyone know you are not patching your fleet… :)

Other than the kernel (and probably some firmware?), what updates require a reboot these days?

It’s not because the reboot requires it (even kernel can be kexec(1) in place for the last 10 years or so).

It’s because your system state is mutable[1] (and so you don’t know if it’s capable of coming back up).

1 - There are distributions where this isn’t the case, but it is for the majority.

I mean, kexec effectively is a reboot, it just plays with the implementation to skip hardware initialization.