On Sun, 13 Nov 2005 05:33:37 +0100, Kevin Raison <[EMAIL PROTECTED]>
wrote:
I am trying to setup a large disksuite RAID5 to handle 100+ million 20KB
files on Solaris 10 x86. However, it seems that UFS has a limit that I
cannot get around. A 1TB filesystem can only have just over 1M indoes.
1.5TB = ~1.7M inodes. I have tried various different arguments to
mkfs, but have yet to find the right value for nbpi. Any suggestions?
Yes. 1 million (or more exactly, 2^20, or 1048576) per TB is correct.
The default nbpi (number of bytes per inode) value for a file system
of > 1 TB is 1 MB. It can range from 1 MB to 2 MB.
The 1 million files per TB limitation is only for file systems of
greater than 1 TB. UFS file systems of less than 1 TB can have a higher
density.
Multi Terrabyte Support for UFS was not intended for use
to host millions of small files but a few large files
like databases.
see mkfs_ufs(1M)
<snip>
mtb=y Set the parameters of the file sys-
tem to allow eventual growth to
over a terabyte in total file sys-
tem size. This option sets fragsize
to be the same as bsize, and sets
nbpi to 1 Mbyte, unless the -i
option is used to make it even
larger. If you explicitly set the
fragsize or nbpi parameters to
values that are incompatible with
this option, the user-supplied
value of fragsize or nbpi is
ignored.
<snip end>
---
frankB
_______________________________________________
ufs-discuss mailing list
[email protected]
