inodes

[Ethan Benson <erbenson@alaska.net>]

I recently noticed that the defaults for mkfs.ext2 have changed 
somewhat recently (or maybe not somewhere after 2.2 kernel was 
finalized)...

the main changes I am interested in are the default block size which 
was 1024 and is now 4096 and the number of inodes created which was 1 
for every 4096 bytes, the new defaults appear to be 4096 block size 
and 1 inode per 8192 bytes, same ratio but you still end up with half 
as many inodes...

having just run out of inodes on my 200MB root filesystem (only /home 
and /usr are farmed out on this system) and having had created that 
filesystem with the older ext2fs utils it has 1 inode per 4096 bytes 
...  (the filesystem has about 77000 inodes which figures about 
right, I don't see anything unusual I am not sure how i managed to 
run out of inodes...)

what is the general opinion on the number of inodes that should be 
made on a filesystem? is there any disadvantage to creating much more 
inodes then default?  (i would guess longer fsck times but that is 
less annoying then running out of inodes...)

also what about the larger block size, I imagine this is faster but 
how much space is really wasted on average by the larger block size?

fortunately I am in the process of replacing this box and the lack of 
inodes is not a huge problem at the moment (i found some files to 
delete so the system can function properly at least), but I want to 
avoid this in the future...

does anyone know what the rationals were for changing these defaults?

thanks


Best Regards,
Ethan Benson
To obtain my PGP key: http://www.alaska.net/~erbenson/pgp/