csync2 for very large file sets

[Terry Gliedt]

Tino Schwarze wrote:
> Hi Matthias et al,
> 
> On Wed, Nov 28, 2007 at 12:40:59PM +0100, Matthias Saou wrote:
> 
>> I have to admit that csync2 is really neat, and replaces in an elegant
>> fashion some ugly rsync+ssh scripts I had been previously using ;-)
> 
> I've been hearing good things about csync2 on this list recently.
> And I've been hearing good things about xfs too. My current situation is
> as follows: We've got a SW-RAID5 with a 900 GB reiserfs file system on it.
> The file system contains BackupPC data of several servers. BackupPC uses
> a pool of files and hardlinks to save space on backups. That is, the
> pool has about 5 million files and 20 million directories.
> 
> Since we've had problems with the SW RAID, the reiserfs suffered and I'm
> about to --rebuild-tree the second time. A new machine (PE1800) is ready
> to take all the data and I'm considering switching file systems since
> I'm not feeling good about keeping this reiserfs any longer.
> 
> My question: Will csync2 be able to handle such an amount of files and
> preserve hardlinks correctly? On another occasion I tried rsync'ing the
> data but aborted after rsync took over 1 GB of RAM while building the
> file list - the machine only has a Gig.
> 
> I'm assuming that xfs will cope well with the data.
> 
> Any advice how to cope with such an amount of data?

We use XFS on any disk larger than 30GB and we have 31TB of disk space 
in use (using various hw raids). The largest disk is 11TB as one 
gigantic /home using XFS.

XFS has worked very well for us. Could some other file system be faster? 
I don't know (or much care) - stability is the important thing to me. We 
have seen a couple of cases where abrupt power failures have left XFS 
confused. In every case running xfs_repair has fixed things up - for 
which I've been very grateful. I'd hope it'll serve you as well. Good luck.