Bug#511180: partman? failure to see ext3 fs on partitioned raid 1 array
> "Leif W" <email@example.com> writes:
> > Currently, mdadm, fdisk and cfdisk, ext2fs utils, and mount utils
> > fully handle partitionable/partitioned raid arrays. I can handle
> > this all with manual commands. But the install scripts do not allow
> > any intervention or injection of manual commands in the partition
> > process.
> Why can't you start a shell and create the partitions, then restart
> partman and continue the installation?
My words may have not been clear. Let me explain. The arrays exist. The
partition (one) exists on the raid 1 array. The filesystem exists on the
array. The data is already on the filesystem. This has been the case for
about 3 years, where my array is has been used as a pure data space, not a
live bootable OS.
The debian install process (partman or whatever is the current set of
scripts), does not recognize any the filesystem on the single partition.
I am not sure if the scripts simply do not see any partitions and just assume
one partition (i.e. if I had 4 and it sees only 1 empty partition). This was
the case in the older bug report #307470 from 4 years ago.
Also I am not sure if it sees my 1 partition, but merely not recognize the
partition type and filesystem (due to using Maximize/NC/Non-Compatible with
MS-DOS) option of cfdisk). Or a combination of both conditions.
> In the end you may also have problems booting the installed system.
> Last time I checked (half a year ago, perhaps) the mdadm initramfs
> integration could not handle partitionable arrays or version 1
> superblocks. Of course things may have changed since then. Or I may
> not remember correctly (my specific reason was different).
I am reading this Linux-RAID Wiki:
Autodetection is deprecated it says. I have set 0xfd partition types for all
RAID partitions. It says only Superblock v0.9 is compatible with autodetect.
I have superblock v1.0. Maybe this is a cause for confusion of the software.
After reading the Wiki, the instructions seem vague, referencing different
"Methods" that do the same basic process, with minor variations using: LILO or
GRUB, raidtab, initrd or initramfs.
It seems I may have to remove 1 partition from the array, repartition,
reformat, install bootable Debian on a normal partition, copy data to degraded
array, configure to boot from the degraded array, and finally add the
install-partition back to the degraded array for the rebuild.
I do not understand why a RAID array that works fine can't be recognized by
the installer. I was hoping to avoid that rebuild work, or to understand and
fix whatever is not working in the install process to make it more robust and
less time consuming for pre-existing arrays.
More research and/or advice needed.