DragonFly users List (threaded) for 2005-08
Re: increase number of inodes?
On Fri, 19 Aug 2005 19:15:30 +0200, Matthew Dillon
:Don't know whether I should report this here, or in Bugs, but here it
:I installed the latest ISO a few days ago. Because I wanted to help
:testing pkgsrc, I cvs-upped the pkg-src tree. I think after about 90%,
:filesystem ran out of inodes.
:I know there is no 'non destructive' way of doing this, so maybe it's an
:idea to increase the number of inodes by default on new installations.
:( It was a small partition (guess it was about 500/600 MB), but maybe
:other people run into the same problems. )
:Good idea/Bad idea?
:Wiger van Houten
:Using Opera's revolutionary e-mail client: http://www.opera.com/mail/
Well, the defaults are pretty good as they stand but they are not
designed for squeezing large numbers of files onto fairly small
partitions. If you wanted to make newfs smarter you could have it
drop back to 1K/8K for filesystems less then, say, 2GB, or have it
mess with the default bytes-per-inode parameter based on the size
of the partition, but it would be a poor heuristic at best.
And changing the block size to such a low value will probably drastically
Just out of curiosity: is it difficult to implement a 'dynamic inode
allocation', such that inodes get created when they are needed?
I read that, for example, HP's VxFS V2 filesystem implements this
This would make file-system tuning unnecessary:
* if you need lots of inodes (e.g. mail server), then you have them (and
small block size)
* if you don't need a lot of inodes, your block size will remain big
enough for good performance
(off course, people running mail servers etc. should know what they're
doing, but it's a good selling point ;) ).
Using Opera's revolutionary e-mail client: http://www.opera.com/mail/