When you lay out your spool file system (the one that holds the actual
articles) with newfs, you'll need to change its defaults to account for
the different ratio of typical file size/number of inodes typically
found on news spool disks.

Articles typically run about 3000 bytes per inode.  The default for a
UFS (BSD fast) file system is 2K bytes/inode which should be sufficient.
But "mkfs" will silently limit the number of inodes in a cylinder group
so some disk formats wind up with too few inodes.  Using smaller cylinder
groups is a solution to this problem.  The following is a typical command
for creating a new spool/news file system:

    newfs -c 8 -f 512 -b 4096 diskname
    
Using a cylinder group size of 8 allows up to double the number of inodes
that the default group size of 16 permits.  Note that using 512 byte
fragments will allow about 20% more articles than the default 1K due to
the small size of the average article.  In any case verify the bytes/inode
is below 3K before using the file system!

Old style Unix file sytems are limited the 65K inodes total.  The
only work around is to create several smaller partitions such that
each file system has <= 3Kbytes/inode.  Then split the news groups
into the various file system to equalize usage as evenly as possible.

© 2008 FAQS.ORG. All rights reserved.