Hi Jeff, You can tune dfs.datanode.balance.bandwidthPerSec in order to change the speed at which the balancer moves data around. It defaults to 1MB/sec so as to avoid using significant cluster resources, but you could certainly bump it up for an individual balancer run if you need it to go quicker.
-Todd On Fri, Dec 18, 2009 at 5:05 AM, Jeff Zhang <zjf...@gmail.com> wrote: > It's my mistake, the balancer process is so slow that make me think it > balance regarding the block > > Jeff Zhang > > On Fri, Dec 18, 2009 at 8:36 PM, Brian Bockelman <bbock...@cse.unl.edu > >wrote: > > > Hey Jeff, > > > > I have not encountered this. > > > > http://dcache-head.unl.edu:8088/dfshealth.jsp > > > > The nodes that are 10x bigger have 10x more blocks on them. What makes > you > > think this is true? > > > > Brian > > > > On Dec 18, 2009, at 4:29 AM, Jeff Zhang wrote: > > > > > Hi all, > > > > > > When I invoke the balancer command, I found that it seems hdfs want to > > > balance the block number of each data node rather than the percentage > of > > > each machine's remaining capacity. > > > Not sure if I am correct, could any help on this ? > > > > > > > > > Thank you > > > > > > Jeff Zhang > > > > >