File locking for simultaneous rsync sessions

[Matt McCutchen]

On Thu, 2008-02-14 at 13:51 -0500, Andrew Culver wrote:
> I have two web servers both mounting the same file system (with GFS). I 
> have a third web server which is independent from the first two, using 
> only its own local file systems. I need to mirror a directory within the 
> shared file system on the first two web servers to the local file system 
> on the third web server. To maintain redundancy, both of the first two 
> web servers will be pushing the files to the third web server via rsync 
> started up by cron as well as on-demand by the web developers.
> 
> My concern is that if rsync is running on the first two web servers at 
> the same time, that there may be a chance for the two rsync sessions to 
> write to the same file(s) on the third web server at the same time, 
> possibly corrupting the files or interrupting the transfers.
> 
> Is there any sort of locking mechanism built in into rsync to prevent 
> this from happening?

Rsync does no file locking.  Multiple rsync processes writing 
concurrently to the destination generally won't make a terrible mess
since each process updates a file by writing the new data to a temporary
file and moving it over the original, but the processes may waste
resources by duplicating each other's work.  If you're using --delete,
the processes might delete each other's temporary files, which could
cause a problem; to avoid that, either protect the temporary files with
--filter=P_.*.?????? or have each process use its own --temp-dir outside
the destination.

> Would running an rsync daemon on the third web 
> server versus running it under an ssh shell make any difference to this? 

The rsync daemon has a "max connections" setting that you could use to
prevent the first and second hosts from pushing at the same time.

Matt