On Mon, Dec 12, 2005 at 11:22:25AM -0600, Brenton Rothchild wrote: > Hi all, > > I was thinking about trying to set up a test cluster > according to the figure in > > http://www.redhat.com/docs/manuals/csgfs/admin-guide/s1-ov-perform.html#S2-OV-ECONOMY > > but I see a post from Benjamin Marzinski that this is a > marketing fluke and isn't possible? > (http://www.redhat.com/archives/linux-cluster/2005-February/msg00049.html) The documentation has been corrected. The picture, as is stands now is a valid configuration. previously, it looked like this ---------- ----------- ----------- | | | | | | GNBD servers ---------- ----------- ----------- /\ /\ /\ A B' B C' C A' Meaning that you were doing mirroring on the gnbd clients. This is untested and unsupported. > Is that true, or this possible? I'd like to make use of two GNBD > export servers that I could RAID 1+0 on the GNBD import nodes > to get a network mirror plus resizeable GFS parition... > if that's even possible. I still believe that it's possible that someone with a bit of time and understanding could hack some stuff and get GNBD running on top of DRBD. I'm pretty sure I explained how I thought that this could be done on a past post, which I can't locate now. I would never do this on a production setup unless you really knew what you would doing, and were willing to support it yourself, and tested it like mad first. The problem with mirroring is that if a machine dies, it may have successfully written to one server and not the other. Now your mirror is out of sync, and a non-cluster aware mirror setup would never know. The non-painful solution is to use cluster mirroring, which will be available through CLVM soonish. -Ben > If anyone could tell me if I'm barking up the wrong tree on this, > I'd appreciate it! > > Thanks, > -Brenton Rothchild > > -- > > Linux-cluster@xxxxxxxxxx > https://www.redhat.com/mailman/listinfo/linux-cluster -- Linux-cluster@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/linux-cluster