Comment 0 for bug 539597

Binary package hint: mdadm

Raid is designed to provide redundancy and failure tolerance. If a raid member fails and redundancy allows it, system operation is not compromised. Upon failure of components the raid array is properly degraded on the fly, and the system *will happily boot* anytime with that degraded array afterwards (without nonsense console droping).

The boot_degraded=NO option (or even default) is therefore an unreliable, disrupting and improper way to notify an admin about a failed raid member. (It is triggered only by raid members disconnected/failing when the system was powered down and compromises operation. It provides by no means a way to prevent the system from running without redundancy and notifying someone about it.)

A better way to notify admins (about all kinds of raid events) is to make the mdadm package depend on a local mail transport agent or wall/notify-send mechanism and stop suppressing the debconf questions for the monitoring deamon.

Please
1) Re-add the init scripts (or new upstart-events) capable to properly start/run necessary arrays degraded during boot (if they don't come up during a timeout period)
2) Remove the bogus boot_degraded debconf option and question.
3) Rely (don't suppress and depend ) on proper means to notify admins about raid events.

/var/log/installer/cdebconf/questions.dat
Name: mdadm/boot_degraded