rsync madness making me crazy

[wagnerone]

I have been reading of all the various rsync versions, spin-offs and variances the past several week and feel like I’ve come no closer to having a “good” solution than I did when I started.
My current situation (which is bound to change soon when I begin updating our servers to Tiger) is an environment of Panther servers with fibre attached Apple raids which I want to sync nightly via the network to Apple raids attached to other hosts. I want these nightly syncs for quick data retrieval (as opposed to lengthy tape retrieval). Just today I was able to restore some data for our production group from my staging sync configuration.
Anyway, currently I have decided to use the aging rsyncX binary to sync the data over the network (using ssh as a transport agent) to the sync host/raids.
So far this apparently is working well… except for one particular sync source which has a hell of a lot of small files. When the sync process gets to this share, rsyncX hits one of its (possibly just an rsync bug from the era of rsyncX build) infamous bugs and hangs.
I can’t use the rsync binary from tiger to do tiger <-> panther syncs.
I can’t use this binary (http://www.quesera.com/reynhout/misc/rsync+hfsmode/) as I need rapid recovery which the sync data resulting from using this binary does not lend itself to. I am syncing HFS <-> HFS.
My options open up apparently when I switch ALL my server hosts to Tiger, but does anyone know of a functional and reliable hfs extended attribute (resource fork) aware rsync that will work via ssh for Panther?

Thanks,
Mike

[Anonymous]

I understand what you’re saying. I have a server running Panther with just one folder containing over 70,000 items in it (and probably another 100,000 items everywhere else) and rsyncX is not thrilled at all… but it does work. Every once in a while it gets some sort of looping malloc error and it doesn’t seem to like files vanishing either. It takes about 2 hours now (after it took 8 hours the first time).

I threw more RAM in the server (a PPC Xserve w/RAID, now 4GB) and the sync machine (dual G5, 2GB) and have success about 95% of the time now. I won’t be getting Tiger until I get a new Xserve, so it’ll probably be Leopard by then. 🙂

j0lt

[Author]

Posts