On Sat, 2008-12-27 at 12:00 -0600, Eric Sandeen wrote:
> On a 256mb filesystem the journal will only be 32mb by default. Still a
> chunk of the fs, but not half! :)
Hmmm, I think this has changed over the years, but it seems the recent
code looks like this:
int ext2fs_default_journal_size(__u64 blocks)
{
if (blocks < 2048)
return -1;
if (blocks < 32768)
return (1024);
if (blocks < 256*1024)
return (4096);
if (blocks < 512*1024)
return (8192);
if (blocks < 1024*1024)
return (16384);
return 32768;
}
It's based on block size. So on a 256mb filesystem, the block size
defaults to 1k, and you get an 8mb journal.
Given /boot mostly stores a handful of kernels, a 4k block size works
just fine and gains a bit of space due to less metadata. (Helpful when
you're trying to maximize space on older systems with smaller drives...)
You then end up with a 16mb journal.
Not as wasteful as I thought, but still, really the journal isn't doing
anything but take up space.
I've been setting up /boot with options something like this on my
machines:
mke2fs -b 4096 -N 128 /dev/sda1
Ah, exercises in obsessive optimization...
Attachment:
signature.asc
Description: This is a digitally signed message part
-- fedora-devel-list mailing list fedora-devel-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/fedora-devel-list
