Is there room for improvement in rescue mode? (was Re: Goodbye, Fedora)

[Andy Green]

Jeff Spaleta wrote:

> For the purposes of this discussion, we will take it for granted that
> at some point in the course of a 3 or 4 releases, many (i dare not say
> most..but many) people who are acting as the primary sysadmin for a
> fedora install will experience some sort of human error which will
> render their system unbootable.  This is an unasailable axiom for the
> rest of this discussion.
> 
> What can we do in the timescale of an F8 release to make using the
> rescue mode easier and more obvious course of action.  Are there ways
> we can advertise its existence as part of sysadmin interaction with a
> normal operating system? Would it be helpful to slip in a rescue
> environment as a grub menu option instead of relying on install media?

That would be cool, another related improvement would be to stick the 
rescue image itself in /boot or in some rescue partition that can be 
defined in Anaconda.  IIRC it was 80MB or something to download the 
rescue-only ISO, that's lost in the noise nowadays for most cases.

> Does it make sense to spend some effort making a more featurefull
> rescue-like environment with guided troubleshooting characteristics?
> What are the top three implementable ideas which would encourage
> casual admins to reach for the rescue environment instead of a full
> wipe and re-install?

I know what would help people "reach for it": if something noticeably 
fatal happens during the boot, reboot into the rescue image 
automatically right away with a path to the broken dmesg and/or 
initscript output sitting there along with some help text.  Maybe even 
use a HW watchdog action to get into the rescue even after something 
really fatal.

Extra bonus power if it can catch the full dmesg / panic somewhere the 
survives the warm boot.

Super invulnerability power points if sshd comes up too if it can find 
the original /etc/sysconfig stuff so even remote servers can be adminned 
after a panic.

-Andy