MORITA Kazutaka wrote: > At Mon, 31 Oct 2011 09:07:25 -0000, > Matthew Law wrote: >> On Mon, October 31, 2011 2:17 am, Miles Fidelman wrote: >>> Something that's not quite obvious from the documentation: >>> >>> How does Sheepdog spread blocks around when nodes have multiple disks? >>> For example, on a 4-node cluster, where each node has 4 drives, is >>> Sheepdog smart enough to replicate blocks in ways that protect against >>> both individual drive failures, and node failures that take out 4 drives >>> at a time? >> Hi Miles, >> >> my understanding of sheepdog is that one would normally run a sheep daemon >> per logical disk. My testing has been by running a single sheep daemon >> per linux md raid 10 device. This is obviously overkill but suited my >> setup and sheepdog's features at the time. >> >> The better solution for current sheepdog is to run one sheep daemon per >> disk and use the failure zones which were recently introduced to tell >> sheepdog 'where' the disks live so it does not place all copies of a block >> of data on a single physical server. > Yes, and Sheepdog uses the same zone id in the same node by default; > dada will be replicated across nodes without any configuration. Thanks guys. (FYI - I currently run DRBD over a pair of nodes that each run md raid 10 - as I go to 4 nodes, I've been looking at different ways of establishing a reliable storage environment that allows VMs to migrate.) -- In theory, there is no difference between theory and practice. In<fnord> practice, there is. .... Yogi Berra |