Hey, In HDFS, the namenode contains difficult to replace data in memory and snapshotted to disk. The secondary takes copies to a different machine.
In HBase, the master doesn't store anything except transient data, and it also manages operations. It can run anywhere, and it will be possible to have a dual master set up at some point. The data is stored in HDFS, so nothing is lost if a master dies. What exactly will be available in 0.20 I can't say yet. The goal at some point is to be fully multi-mastered with ZooKeeper coordinating our primary-master selection. On Thu, Apr 16, 2009 at 5:13 PM, Jae Joo <jaejo...@gmail.com> wrote: > Is "Secondary Name Node" in hadoop can be used in hbase or is it what we > can > expect in 0.20? > > Jae > > On Thu, Apr 16, 2009 at 7:38 PM, Jae Joo <jaejo...@gmail.com> wrote: > > > I have read and foind that the master node is SPOF and 0.20 will have > > failover system for master node. Any update or any solution to avio > > SPOF? > > Jae > > > > On 4/16/09, Billy Pearson <sa...@pearsonwholesale.com> wrote: > > > hbase and hadoop both have a single node (namenode & master) that tell > > the > > > clients where stuff is that's all they do is keep up with > > > where stuff is they do not handle the getting of the data the client > api > > > will go to the node with the data they tell it to find the data and get > > the > > > data. > > > > > > as for adding new nodes to hadoop to balance the cluster you can use > > > bin/start-balancer.sh -threshold 5 > > > Where 5 is the % you want the nodes to be with in each other in terms > of > > > used % > > > > > > Billy > > > > > > > > > "Jae Joo" <jaejo...@gmail.com> wrote in > > > message > > > news:3eb24ef50904161430we261882v12600704f521...@mail.gmail.com<news%3a3eb24ef50904161430we261882v12600704f521...@mail.gmail.com> > . > > .. > > >> Hi, > > >> > > >> We are doing some feasibility study of HBase and have questions. > > >> 1. Once we have build tables on certain number node and decide to add > > more > > >> nodes, how it is working? I know that it is not hbase, but hadoop > > >> questions. > > >> Will it automatically redistribute the files or start utilizing the > new > > >> node > > >> for new inserted records? > > >> > > >> 2. Is there any way to have multile master nodes? Should all traffis > (or > > >> requests) be going to SIngle Master node to access the data from > hbase? > > >> > > >> > > >> Thanks, > > >> > > >> Jae > > >> > > > > > > > > > > > > > -- > > Sent from my mobile device > > >