How to create filesystem with lower reserve space

[dictum9]

11.23

I know by default, the reserve space is 10% but I want it to be 1%

What option do I use?

[Patrick Wallek · Accepted solution]

What type of file system are you creating?

HFS
VxFS
Something else?

For information check the man page for the newfs and/or mkfs command for the file system type you want to create.

For example:

man newfs_hfs
man newfs_vxfs
man mkfs_hfs
man mkfs_vxfs

[dictum9]

-F vxfs

[Pete Randall]

I, at least, am going to need more information. The only way I know of to create a file system is by using the newfs command and the newfs command says absolutely nothing about reserve space.

What are you referring to? Which command? What type of "filesystem"?


Pete

[James R. Ferguson]

Hi:

I would assume that you are referring to the 'minfree' value for HFS filesystems. This is set during a 'newfs' but can be changed with 'tunefs'. HFS filesystems are deprecated.

Regards!

...JRF...

[Pete Randall]

OK, vxfs. However, I know of no "reserve" in vxfs so I'm still confused. Can you be more specific?


Pete

[dictum9]

These have higher reserve space (10%)

# fstyp -v /dev/vgdevsap/lvol70
vxfs
version: 5
f_bsize: 8192
f_frsize: 1024
f_blocks: 18432000
f_bfree: 18410945
f_bavail: 17260261
f_files: 4603024
f_ffree: 4602736
f_favail: 4602736
f_fsid: 1073807429
f_basetype: vxfs
f_namemax: 254
f_magic: a501fcf5
f_featurebits: 0
f_flag: 0
f_fsindex: 9
f_size: 18432000


these have lower reserve space of 1%

# fstyp -v /dev/vgdevsap/lvol51
vxfs
version: 5
f_bsize: 8192
f_frsize: 1024
f_blocks: 13631488
f_bfree: 540014
f_bavail: 506264
f_files: 135544
f_ffree: 135000
f_favail: 135000
f_fsid: 1073807410
f_basetype: vxfs
f_namemax: 254
f_magic: a501fcf5
f_featurebits: 0
f_flag: 0
f_fsindex: 9
f_size: 13631488
sapdev01:/#


How do I adjust this reserve space, nothing in the man page on newfs mentions it.

[James R. Ferguson]

Hi (again):

> How do I adjust this reserve space, nothing in the man page on newfs mentions it

That's because there is no way to do this for a VxFS filesystem. A filesystem reorganization (with 'fsadm') may change the value you calculate but that's it.

Regards!

...JRF...

[dictum9]

I don't see any such option in fsadm either.

[Patrick Wallek]

I don't think JRF didn't mean that you could **change** the value by using 'fsadm'. I think he meant that the value that is calculated may change after reorganizing the filesystem with fsadm.

In looking at your fstyp output, I don't understand where you are getting the 10% & 1% reserve figures.

[James R. Ferguson]

Hi (again):

I think what you are seeing it the difference in the average file sizes in each of the two filesystems.

If you look at the values for 'f_blocks' and 'f_files' given the same value for 'f_bsize' you will see that for the first filesystem each node (file) consumes about 4 blocks on average. For the second filesystem, each node consumes about 100 blocks.

A file whose actual size is _less_ than 8K (8192) still has to use the lowest unit of space --- the 'f_bsize'. Hence for the first filesystem I would expect a higher proportion of wasted space.

The block size is set during creation and ultimately governs how large the overall filesystem can be:

http://www.docs.hp.com/en/B2355-60105/mkfs_vxfs.1M.html

Regards!

...JRF...

[OldSchool]

I think that, by "reserve space", hes referring to the space root could get to in the event you got a "disk full" error....

as in somebody fills / or whatever causing logins to stop. root can usually get in because the default was to leave 10% available (which might include space for the inode tables for all I know).

i didn't realize that was "tunable" however

[Patrick Wallek]

OS -- I agree with what you say about the min free space.

>>i didn't realize that was "tunable" however

It is on HFS file systems. The '-m' parameter to newfs / mkfs tunes that. It's not tunable on VxFS though.

[Steven E. Protter]

Shalom,

I would suggest backing up the data and running newfs again is the only route available to you.

Since fsadm can not tune it, that appears to be the only way to go.

I might first though look to see if lowering the reserve space to that low a level might not cause performance impact as the filesystem gets closer to full status.

SEP