hardlinks

[Dan Lund]

Darrin Chandler wrote:
> Hardlinks are what you have all the time. The only difference is that
> normally there's only one of them. They are not a different type of
> beast, in the way that symlinks are.
>   
Yeah, I know what a hardlink is, basically a "file" with a pointer to 
the same inode as the first file, along with an incrementing of the 
link-number on that inode.

> Two options that come to mind...
>
> 1. Write a script to prune. You'll want to be examining the actual
> directory entries and comparing vnode/inode stuff to see if a given file
> is the same actual file. Start with oldest subtree, read in all dir
> info into "A". Read dir info for next oldest subtree into "B". Unlink
> duplicates on disk indicated by "B", but leave the info in the program.
> Set "A" equal to "B". Read next oldest "B". Loop at compare. Wash.
> Rinse. Repeat. This is not rocket science, but it involves actual
> coding, and it's not something you want to get wrong with your real
> backup data. If you go this route, make some fake stuff and test
> thoroughly!
>   
Something of that magnitude would take multiple gigabytes of memory, and 
many many days.  I was going to go that route, but unfortunately I'm 
screwed in that respect.

> 2. Use software such as dump, which backs up the actual filesystem
> structure, and as such, will not use any more space. Then you can put
> your dump on tape.
>   
I'm looking into this route right now... just trying to find 1.7Tb of 
diskspace free.

I just contacted CommVault technical support and they said that the 
hardlinks are backed up as actual files and restored as links to 
preserve the integrity of the data on the backup.
I call BS.