Then, I suggest: 1) Increase your session per user in RAC equal or greater than your map tasks 2) I suppose you want 100 map tasks, then write a little java code to generate 100 files, each file has only one line, you can write any in this line. 3) Define a STEP for all maps. then each map task select STEP records from DB. 4) Put all your files to the HDFS 5) Create Map(no reduce here), execute "select * from Oracle where rownum >= lineno and rownum < maptask_id * STEP" in each map task.
On Fri, Dec 14, 2012 at 11:36 AM, Mehmet Simsek <[email protected]>wrote: > Hello Azuryy, > > Oracle in RAC.because of too long running sql statement datanode cannot > respond to namenode about statistics.because of that exception is thrown by > namenode like as "cannot respond from data node in 600sec..." In sqoop tool. > > Is there an another solution about this problem? > > M.S. > > On 14 Ara 2012, at 05:11, Azuryy Yu <[email protected]> wrote: > > > Hello Mehmet, > > > > what about your oracle? a single node or RAC ? > > > > > > On Thu, Dec 13, 2012 at 11:46 PM, Mehmet Simsek < > [email protected]>wrote: > > > >> Hi,I want to load 5 billion rows from oracle table to hbase table. Which > >> technique is the best for this bulk loading? > >> > >> Thanks > >> > >> M.S > >> > >> >
