df reporting incorrect size?

I'm a linux-lvm newbie (some limited use of IBM's LVM in AIX), and some use of Veritas in Solaris. So I'm doing some testing, hoping to find out that lvm will serve my needs and that of my employer.

I've done all of my fdisks to change partition type (0x8e), vgscans, pvcreates, vgcreates, lvcreates, mke2fs on the lv etc, (used nice generic names for my testing, like vg and lv), mounted it up and get the following:

mount /dev/vg/lv /test
df -h /test
Filesystem Size Used Avail Use% Mounted on
/dev/spheravg/spheralv

                             59G   20k   56G   0% /test

WTF?!?! Where'd that 3G go? Does LVM mess with things so that the OS cant read volumes correctly? Or is there another issue that I might be running into?

Thanks in advance for feedback/assistance/HELP.

Also, on another note, how difficult is it going to be for me to add another ~63G to this LV from another physical device? The whole reason we're looking at LVM is mainly to perform this function (one HYUGE filesystem over multiple devices).

This isn't a LVM question, it's a filesystem question.... Specifically:

$ man mke2fs
...
       -m reserved-blocks-percentage
              Specify the percentage of the filesystem blocks reserved for the super-user.
              This value defaults to 5%.
...

So 5% of 59GB is ~3GB. 59-3 = 56GB available.

As you can tell, 5% is huge. For non-root/OS filesystems, I almost always set that to 0. Look at tune2fs to see how to do that without recreating the FS.

Randomly Generated Tagline:
Real theology is always rather shocking to people who already
 think they know what they think.  I'm still shocked myself.  :-)
              -- Larry Wall in <[email protected]>

My only advice is that you might want to reserve a small number of blocks
for root on the filesystem, read up on -r for tune2fs. The percentage as nearly as I can tell is only in integers.

If you are using ext3 (which isn't a bad idea for a filesystem this
large), I've used ReiserFS on filesystems this big. The journal will take up something between 1024 Blocks and 10240 blocks depends on how you set it up, which can use up an additional 400M. Read up on the -J option to tune2fs. This can be reset after the filesystem is configured.

If you're going to all that trouble, you should consider how many files
you might put on the filesystem, and adjust the number of inodes to be reasonable. Look at -i and -N for ext2/3 (this can only be done at filesystem creation time). df -i will report on inodes.

Finally, adding more space to the filesystem:

pvcreate /dev/device
vgextend VolGroupName /dev/device
lvextend --size +60G LogicalVolumePath

Then depending on the filesystem you have to resize it. To do that on
ext2/3, you have to unmount the filesystem. For reiserFS, you can do it online.

If all you really wanted was a huge filesystem of a known (fixed) size,
and you aren't planning on adding any drives, the md tools and software RAID will do the job (either linear append or RAID 0). They aren't nearly as flexible, and aren't half as cool in my opinion, but they do get it done.

It can save you a lot of trouble w/ LVM to not use it. LVM if you need
it is a godsend, if you don't I wouldn't bother with it, it's can be a hassle. I play with it, so I get used to it. It has caused problems in the past on my personal boxes. I haven't had catastrophic stuff happen to me, and the guys here appear to work miracles on data recovery. So, I wouldn't be paranoid about my data, but downtime is a problem for me, so I skip all the features I can do without on my production machines at work.

Thanks,
Kirby

3G = approx 5% of 59G filesystem, or the reserved blocks on a standard ext2fs filesystem.

They're right there where you expect to find them if you use a unix filesystem.

<quote who="Ben Snyder">