> Ok, I finally answered one of my questions when running the rebalance > tool. One node is 180gig and another 20gig so it technically in balance > with a 10% ratio. When I ran 1%, it did rebalance. I guess I expected > hbase to be writing to both nodes but was only writing on one. I need > to get to a more equalized cluster first to run more tests.
Do you have a region server on each node? If not, and replication is set to 1, then the blocks will be assigned to the closest datanode to the client so in this case the local node. > QUICK QUESTION: Is there a quick way to limit a node to the size of data > it will use(to lower that 180gig number to 20)(or do I just have to > create a new linux partition with new mount point and point that node > there)? Different partition is one thing, you could also set a special dfs.datanode.du.reserved for the bigger disk... but you'll probably spend more time dealing with the heterogeneous nature of your cluster than if you just used the node with 180GB and the result should be the same. J-D
