Bug#383612: mdadm: upgrading to 2.5.2-7 broke my array, removing mdadm stuff from initramfs fixed it

[Lars Wirzenius]

Package: mdadm
Severity: normal

I'm not sure what the severity should actually be.

The machine has a four disk raid5 array and runs testing. Yesterday
I upgraded it to current testing. It had been a while since the previous
time. If I read the dpkg.log correctly, mdadm got upgraded from 2.4.1-6
to 2.5.2-7.

Today, when I booted the machine, my array wouldn't start. After fiddling
about (trying various versions of mdadm), what finally worked was
2.5.2-6, followed by dpkg-reconfiguring it so that it won't start any
raid arrays in initramfs (the previous debconf answer was "all" for the
"start raid for root fs" question; I replaced it with a blank answer).

This fixed it.

This makes me think there is something wrong with the way an upgrade
happens, or in the way the stuff that gets put into the initramfs works.

If I can provide further details, don't hesitate to ask.

(Oh boy, one gets nervous when debugging the failure of a 680 gigabyte
array with all your life on it. It's not the only copy, of course, but
the thought of having to restore things from DVDs is going to give me
nightmares for *weeks*.)


-- System Information:
Debian Release: testing/unstable
  APT prefers testing
  APT policy: (500, 'testing')
Architecture: i386 (i686)
Shell:  /bin/sh linked to /bin/dash
Kernel: Linux 2.6.17-1-486
Locale: LANG=fi_FI.UTF-8, LC_CTYPE=fi_FI.UTF-8 (charmap=UTF-8)