Subsequent reboots after entering recovery mode will sit at the grub menu forever

[jdreed]

Thanks to a grub config that thinks it's smarter than you, once a machine has been booted into recovery mode, subsequent reboots will sit at the grub menu waiting for input until the machine has successfully booted _AND_ cleanly shutdown at least once using the normal kernel. We should figure out which grub option to kick to fix this. There is a GRUB_TIMEOUT set in /etc/default/grub, so I'm not sure what we're missing here. GRUB_HIDDEN_TIMEOUT, however, is set to 0.

[geofft]

This is from /etc/grub.d/00_header:

cat << EOF
if [ \${recordfail} = 1 ]; then
  set timeout=-1
else
  set timeout=${GRUB_TIMEOUT}
fi
EOF

The easiest way to set another timeout is to drop a later .d script that prints a different "set timeout" line. For -cluster, unconditionally using $GRUB_TIMEOUT is fine. For -workstation or below, I'm uncomfortable messing with GRUB, but if we have to, copying the code in a later .d script using something like 300 instead of -1 would at least let the machine boot unattended. (But usually if you're going to the recovery menu and not exiting the recovery menu by booting cleanly, it's because the machine won't boot normally and needs manual attention, so this behavior is reasonable.)

Depending on what you want to do here, e.g. a recovery menu item that automatically reboots, running grub-editenv /boot/grub/grubenv unset recordfail (cf. /etc/pm/sleep.d/10_grub-common) just before rebooting would also work.

[jdreed]

I'm only considering cluster.

[jdreed]

r25234

[jdreed]

c-l-c 1.34 ->dev