Hi, The balancer works with the average utilization of all the nodes in the cluster - in your case it's about 13%. Only nodes that are +/- 10% off the average will be rebalanced. Node 4 isn't under-utilized because 13-10=3 which is less than 4%. You can use a different threshold than the default 10% (hadoop balancer -threshold 5). Read more here: http://hadoop.apache.org/core/docs/current/hdfs_user_guide.html#Rebalancer
Tamir On Mon, Apr 27, 2009 at 11:36 AM, Usman Waheed <usm...@opera.com> wrote: > Hi, > I had sent out an email yesterday asking about how to balance the cluster > after setting the replication level to 2. I have 4 datanodes and one > namenode in my setup. > Using the -R switch with -setrep did the trick but one of my nodes became > under utilized. I then ran hadoop balancer and it did help but upto a > certain extent. > > Datanode 4 noted below is now up to almost 5% but when i try to balance the > datanode again using the "hadoop balance" command it says that the cluster > is already balanced which isnt. > I wonder if there is an alternate way(s) or maybe overtime Datanode-4 will > pick up more blocks? > > Any clues? > > Thanks, > Usman > > Name: 1 > State : In Service > Total raw bytes: 293778976768 (273.6 GB) > Remaining raw bytes: 222235858599(206.97 GB) > Used raw bytes: 48140136448 (44.83 GB) > % used: 16.39% > Last contact: Mon Apr 27 08:34:46 UTC 2009 > > > Name: 2 > State : In Service > Total raw bytes: 293778976768 (273.6 GB) > Remaining raw bytes: 231235100994(215.35 GB) > Used raw bytes: 40704245760 (37.91 GB) > % used: 13.86% > Last contact: Mon Apr 27 08:34:45 UTC 2009 > > > Name: 3 > State : In Service > Total raw bytes: 293778976768 (273.6 GB) > Remaining raw bytes: 211936026161(197.38 GB) > Used raw bytes: 59591700480 (55.5 GB) > % used: 20.28% > Last contact: Mon Apr 27 08:34:45 UTC 2009 > > > *Name: 4 > *State : In Service > Total raw bytes: 293778976768 (273.6 GB) > Remaining raw bytes: 258876991693(241.1 GB) > Used raw bytes: 12142653440 (11.31 GB) > % used: 4.13% > Last contact: Mon Apr 27 08:34:46 UTC 2009 > >