On 01/31/09 09:44, Keith Freedman wrote: >> Well, I can see a couple of cases where it would be expedient to >> change the configuration. Lets say one of the machines losses a disk >> and the machine won't be fixed for a week say. Then, once the machine >> comes back up, I could use a different configuration and sync the >> files locally on the machine that is available rather than do it over >> the slow network. I am guessing that this should work. > > well. it depends I suppose. where's this data coming from that you're > syncing? From the AFR'd copy on the other machine. > if it's in a backup, this would work, as long as the backup restores the > extended attributes as well. > then run ls -lR and it will just catch everything up. Right. > you could also do this by setting the read-subvolume to the > fastest/closest server. [...] > depending on the volume of data and speed of your network, it might be > just as fast or faster to just let gluster copy over everything, rather > that restore then auto-heal. Well, the network is slow (10Mbps on average) so if there is a large amount of data, the best thing to do would be to bring the disk over, mount it, rsync/cp -a (or use a modified gluster client/server setup) and take it back followed by an ls -lR. cheers, prabhu
