mke2fs options for very large filesystems
tytso at mit.edu
Wed Feb 9 02:03:22 UTC 2005
On Tue, Feb 08, 2005 at 12:59:53PM -0700, Andreas Dilger wrote:
> On Feb 08, 2005 08:00 -0700, kwijibo at zianet.com wrote:
> > Just as a side note but do you really want to reserve 5%
> > of your 2TB partition for root?
The filesystem performance will degrade if you eat into the last 5%,
but as long as you're willing to live with this tradeoff, and assuming
you don't need to reserve space for things like log files, then sure,
you can always drop the percentage.
> > Another side question on this. Is it possible to make mkfs
> > accept non-integer values for the reserved percentage?
> > It is getting to the point these days where partitions
> > are getting so big that even 1% is becoming quite a waste.
> > However it would still be good to have some space reserved.
> The value is stored in the kernel as a blocks count, so it is possible
> to store any value there. It should be possible to change the mke2fs
> and tune2fs option parsing to accept a float there. I think you can
> set this manually via debugfs and "ssv" (in the latest e2fsprogs there
> is also a command to set a single superblock value).
The officially supported to set a specific number of reserved blocks is
tune2fs -r <number_of_reserved_blocks> <filesystem-device-file>
Since mke2fs sometimes is used in install floppies, etc., I decided
there was no point to drag in floating point routines, especially when
you can always set the number of reserved blocks directly using
tune2fs -r after you create the filesystem.
More information about the Ext3-users