[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

RE: Initramfs-tools forcefully stars md arrays. How to disable it?

> On Sun, 5 Aug 2012 22:01:49 -0700 <blac1175@gmail.com> wrote:
> initramfs-tools (0.98.8, the one in squeeze) uses the command below to
> start up the array for the root filesystem (from line 91 of
> /usr/share/initramfs-tools/scripts/local-top/mdadm):
> $MDADM --assemble --scan --run --auto=yes $dev
> This specifies --run, which makes mdadm start the array even if it can
> only do so in degraded mode. This means, if I knock a SATA cable
> loose, my machine will still boot, but dirty the array. Once I
> notice, and reconnect the cable, I now have to run risky mdadm
> commands to add the drive back, and then I have to rebuild the whole
> array.
> What I'd actually like, is if my system failed to boot if any drive is
> missing from the array. Then I'd have the opportunity to check the
> cables, and if needed, boot a live CD for troubleshooting. I'd like
> to make my own determination when do I give up on a harddrive, and
> bring the array up in degraded mode. I understand, that probably for
> most people, this is not what they want. They want high availability,
> that's why they use raid. Maybe I'm using raid in the wrong
> situation? Anyways...
> My question is how can you control this behavior? I see no way of
> configuring initramfs-tools not to add the --run option. Would it
> make sense to have such an option somewhere in the configuration?
> Also, could we take this as far as giving a prompt if the array did
> not start in normal mode, to let the admin force the start in degraded
> mode right there, without any need for a live CD or other maintenance?
> I don't know how easy is it to do anything interactive from
> local-top, and how portable that would be.

Perhaps you can override the default behavior  with a md file/script in your initramfs.

if [ -e /scripts/local-top/md ]; then
  log_warning_msg "old md initialisation script found, getting out of its way..."
  exit 1


Reply to: