Re: partman: please create ext3 file systems with online resize support

[Goswin von Brederlow <brederlo@informatik.uni-tuebingen.de>]

Petter Reinholdtsen <pere@hungry.com> writes:

> [Goswin von Brederlow]
>> 100GiB is either way to high (for my 200MB / partition) or way too
>> low (for my 1TiB data partition). I think the value should be less
>> than 16TiB (maximum size) and no more than 10 times the filesystem
>> size.
>
> Yeah.  I read from the mke2fs manual page that -O resize_inode set the
> max-size to 1024 times the current size.  That might be an OK default
> for ext2prepare as well.  Perhaps it should be patched to make that
> argument optional, and use 1024 * current size in that case.  Patches
> most welcome. :)

Doesn't that waste a ton of space for resizing or is that just a minor
part in the superblocks? If the size impact is so small then I retract
my 10* limit.

>> It would probably be worth a patch to mke2fs to output in D-I
>> progress format though instead of parsing the current output and
>> translating that.
>
> Yes.  But I am not sure if such patch would make it into etch in time. :)

Well, last week/month would definetly have been better.

>> But all of this won't remove the need for ext2prepare. Afaik there
>> is no option for mke2fs to build a filesystem already perpared for
>> resizing.
>
> Are you aware of the -O resize_inode option to mk32fs, currently the
> default for mke2fs?

I am now. Thanks.

MfG
        Goswin