New file system installed

[Sanjay Tailor]

Hello,
I just added a new disk using pvcreate and then used vgcreate and then lvcreate to set up a new but unused disk. I have NO files in this lv and it is mounted. When I do a BDF command I get the following output for that lv:
Filesystem kbytes used avail %used Mounted on
/dev/vg02/lvol1 2003481 9 1803123 0% /facts

Looking at this it seemd as if 9kb is being used. It tells me that %used is 0%. But if you look at kbytes and avail columns it is very strange. 2003481 kbytes with 1803123 avail. That looks like 200MBs of space. Is this space being used for something?

Any ideas? Is this the LVM overhead but the manual said it would vary up to 400kb?

Any help would be appreciated.

Thank you.

Sanjay Tailor.

[Paul Hite · Accepted solution]

You are seeing the result of minfree which is typically set to 10%. This is to ensure that any files allocated do not become fragmented. Root can continue to use space after after the disk is full. If a root process fills up a filesystem, bdf will show the disk as 111% full. In this case, you can be sure that the last file written was badly fragmented as it had to suck up all remaining space.

Take a look at "man tunefs".

[William Dy]

Please note too that "tunefs" only applies to HFS type filesystems. Normally, the default of this is 10% of total fs space, but you can tune it all the way down to 2% assuming that it contain databases. This would give you more room to put files. If you're using JFS or VxFS, there's not much you can do with this overhead.

[Bill Hassell]

Space is needed on any filesystem for the directory structures. For HFS, there are preallocated cylinder groups and super blocks. VxFS has a small number of structures but these are built on the fly. minfree can be any value including 0. If set to zero, then as the filesystem nears 90% full, finding new space will take a significant amount of time. The minfree value also allows root to login and perform tasks on a 100% full filesystem (/ should have minfree set to about 5-10% for safety).

For HFS, you can reduce the overhead by increasing the block size from 8k to 64k with newfs. This is ideal when you have few files but they are large in size.