inodes taking that much space on FreeBSD?

Portal Home > Knowledgebase > Articles Database > inodes taking that much space on FreeBSD?

Posted by Lem0nHead, 07-20-2008, 04:17 PM
hi I just noticed something weird on my FreeBSD server FreeBSD Disklabel Editor shows: twed0s1e /home 108GB * df -h shows: /dev/twed0s1d 105G 4.0K 96G 0% /home so... 108 GBs become 105 GB that become 96 GBs usable? that's 11,11% loss pretty big IMO (this has nothing to do with "real space"... it's really GBs (1024^3, not 1000^3)
Posted by HoundOfTheSmith, 07-20-2008, 04:29 PM
As has been mentioned in other posts, and covered in the newfs man page, the BSD OSs reserve a particular (tunable) percentage of disk space for root only.
Posted by The Ally, 07-21-2008, 06:01 AM
Lem0nhead, you want to read the output of "man tunefs", in certain the "-m" parameter. Short form: 8% of file space are by default held back from normal users. You can reduce that, which can have impact on the performance. Anyway, on a /home partition, I cannot see many reasons why you shouldn't reduce it to 0% (whereas on "/var" and "/" you would NEVER want to reduce it to 0%, as you will not be able to login via SSH any more when even root cannot write to the disk). Cheers Carsten
Posted by zuborg, 07-21-2008, 06:13 AM
The reason is that overfilled FS works very slow. It's too expensive to find not used blocks, they are not optimal to store data for fast access in future, file fragmentation level grows significantly... The optimal minimum level of free space varies from case to case, but 8% is quite common value to keep system away from overfill suffering - I really don't recommend to lower this tunable.
Posted by The Ally, 07-21-2008, 06:21 AM
Yes sure. What I meant is that as an absolutely last resort it could be done on a partition where root does not necessarily needs to have a reserve. Of course, it is always better to have a reserve left and to free space than to use every single available byte :-) My initial posting might have been a bit misleading...
Posted by Lem0nHead, 07-21-2008, 11:12 AM
got it thanks for the replies I think I'll leave it as it is, since I i/o is a important bottleneck and this would make things slower

Add to Favourites Print this Article