initramfs/udev/mdadm/lvm2 integration

Jerome Haltom wasabi at
Sat Sep 15 16:15:08 BST 2007

Another aspect of this I think is we don't really handle notification of
an array failure very well on the desktop. If it boots degraded, on my
desktop machine, I doubt I'd ever even notice. mdadm would email root.

We need it to move through HAL and result in a notification pop up.

On Fri, 2007-09-14 at 09:58 -0700, Kees Cook wrote:
> Hi,
> On Fri, Sep 14, 2007 at 11:17:13AM -0400, Phillip Susi wrote:
> > Kees Cook wrote:
> > > So, the question is, which is the "least surprising"?
> > > 
> > > a) having the machine spew "I can't mount, here's what you can do...",
> > >    potentially endangering SLA and/or convenience.
> > > b) mounting degraded, possibly due to poor timing, potentially endangering
> > >    partitions and/or data.
> > > 
> > > I personally think "b" is more surprising.  I don't think it would be
> > > hard to add another boot-time flag that means "auto-boot-when-degraded"
> > > (which could be mentioned in the spew from "a").
> > 
> > I find A to be a lot more surprising.  The whole purpose of having a 
> > raid 5 setup is so the system will continue to operate just fine in the 
> > event of a drive failure.  This includes booting up, which is often when 
> > failures occur.  Sure, needlessly degrading the set isn't good, but the 
> > system does what it was designed to: keep running.
> This is, I think, where we may need to turn to the tech board to get
> a decision.  Unless Scott trumps us  ;)
> I wonder if specifically asking the question during mdadm install would
> be another way to handle this?  Make it a debconf question that affects
> the "default" initramfs behavior?

More information about the ubuntu-devel mailing list