[Yaird-devel] Bug#351183: --run does not fix the disk renaming problem

[dean gaudet]

[i've cc'd the md/mdadm maintainer so he can chime in if i'm making an 
error... Neil if you want to see the entire thread it's visible at 
<http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=351183>.]


On Tue, 2 May 2006, Jason Lunz wrote:

> I agree. Unfortunately, telling mdadm to scan other devices/partitions
> is only useful if there will be device nodes for them. And by default,
> yaird only adds the dev nodes that are listed in the above command.

please take a look at the README.initramfs file which appears in unstale
mdadm packages ... (such as the 2.4.1-5 that's now in unstable).  it says
the following:

  We do not need to create device files for the components which will
  be assembled into /dev/mda.  mdadm finds the major/minor numbers from
  /proc/partitions and creates a temporary /dev file if one doesn't already
  exist.

this is why "-Ac partitions" will work...  Neil documented this in
detail only recently, but the same is true of older mdadm packages.
Neil documented this partially as a result of dealing with confused debian
users on linux-raid mailing list...  because debian's initrd-tools,
initramfs-tools and yaird all have the same/similar problem --
see <http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=338200>,
<http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=361674>.


> So it's useful to add --run, so that the --assemble command will start
> an array that's missing a disk. But telling it to scan partitions with
> --config only has the effect of segfaulting sarge's mdadm, or making
> sid's mdadm behave more or less as it would with --run.

i'd be totally happy if the fix appears only in etch.  if you really want
it to appear in sarge (i assume you're thinking of some planned bugfix
release) then i'll go start petitioning the mdadm maintainer to upgrade
to the last 1.12.x in sarge.  but we need to fix the bug somewhere and
i'm happy to start with etch and yaird...


On Mon, 29 May 2006, Brendan Cully wrote:

> So I vote heartily to add --run to the yaird mdadm template. Patch
> attached for your convenience.

i think --run is a bad idea... the docs say nothing about it helping to
start a degraded array:

	Insist	that  mdadm run the array, even if some of the components
	appear to be active in another array or filesystem.  Normally
	mdadm will ask for confirmation before including such components
	in an array.  This  option  causes that question to be suppressed.

i'm skeptical that at this point in boot any of the components would be
in use anywhere else since we're trying to construct the root filesystem.

Brendan did you test this patch?  from your description it sounds like
you had fixed your problem from a rescue disk before developing the patch...

fwiw i'm actively running the patch i originally posted... and i've
suggested the use of yaird and my patch to many users on linux-raid who
have had device renaming issues on unstable/etch -- and it's working
for all of us...

thanks
-dean