Back to RequestsForEnhancement
2004 August 10 KeithLofstrom
A backup disk for dirvish may grow to over 100 images; the hardlinked copies of images will use a lot of inodes. This can fill a disk; for over 100 images, use 4K data per inode instead of the default 8K data/inode!
This makes all the standard copy procedures:
cp -a oldfiles newfiles ( cd oldfiles ; tar cpf - . ) | ( cd newfiles ; tar xpf - ) rsync -a oldfiles newfiles
... very, very slow. Like around 10 GB/day . Yikes!
While you can do a "dd" if the old and new partitions are the same size, this does not help if you are changing partition sizes or data/inode ratios.
The proper solution will probably be a partition copy tool that works below the file system virtualization layer, directly with libext2fs . I don't know how to write that.
Comments? Please enter data and WikiName.