[mdlug] More new hard drive details

[Aaron Kulkis]

Jeff Hanson wrote:
> On Thu, Jan 15, 2009 at 2:07 PM, Aaron Kulkis <akulkis00 at gmail.com> wrote:
>> That's all well and nice, but it doesn't tell us much.  The crucial
>> piece of the pie is the LVM configuration -- even if NONE of your
>> non-root partitions gets corrupted, if the LVM config is corrupted,
>> then the rest of the partitions are.... recoverably ONLY with the
>> investment of significant time and $$$.
>>
>> As for the filesystems themselves, they don't care whether they
>> are on an LVM or not... the filesystem is just a string of bytes,
>> whether on one contigous partition, or in blocks added willy-nilly
>> from several block devices.
> 
> My root partiton was corrupted as the memory error caused the kernel
> to write past the end of the disk (and probably random writes
> everywhere else).  LVM2 tools just look for the partition metadata,
> the config files are only a convenience:
> http://dev.riseup.net/grimoire/storage/lvm2/#how_to_use_a_lvm_devices_on_a_different_machine_
> 
> How many LVM2 recoveries have you performed and what problems did you encounter?

Fortunately, none so far.
But I've been dealing with LVM systems from various vendors
since the mid 1990's and the gain in flexibility generally comes
at the expense of putting another link in a chain of things
which must NOT have any errors.

Where is the current Linux LVM putting the LVM data?
I can't see it being stored in the boot sectors, because there
just plain old isn't enough space there.