I use rsync running on MacOS X to mirror some remote Linux file systems to an externally connected Firewire drive (“Mirror”), and to archive changed and deleted files to a second externally connected Firewire drive (“Archives”). In general, it works fine, but there has been a long-standing tricky problem, that my system administrator (Niall O Broin) and I finally solved today. Since I couldn’t find anything about this in Google, I figured a blog post was in order.
Here’s the basic problem: Every single time I run the rsync command, a certain set of files are *always* viewed as changed, and subsequently downloaded (and archived.) And I’m talking about static files like JPG images, not constantly changing stuff like log files.
Well, we finally nailed it. Have a look at this directory listing from the Linux server:
-rwxr-xr-x 1 webrun www 42648 Nov 7 2005 Uirapuru.jpg -rwxr-xr-x 1 webrun www 67248 Nov 7 2005 uirapuru.jpg
The problem is that the Linux filesystem is case *sensitive* — you can have two files with the same name, if the case of their names is somehow different (as above). However, on most MacOS X formatted drives, the file system is case *insensitive*.
So, every time rsync would run, it would compare the *two* Uirapuru files with the *single* local file, and, of course, one of them would be different, and get downloaded (and subsequently sending the alternate file to my archives.)
What’s the solution? The solution for me is to reformat my Mirror and Archive drives to case-senstive HFS+. (The problem, though, is I don’t have the additional space to copy my files temporarily, in order to reformat the volumes. So, I guess I’ll just live with this for a while…)