Hi James,

I noticed the same thing and thus mentioned the server having 2 ethernets.
No I do not have any hbase or hadoop configuration in my classpath. Any
ideas what might be causing the problem?

Regards,

Imran

On Oct 7, 2010 2:06 PM, "James Baldassari" <[email protected]> wrote:

Hi Imran,

Glad that fixed your XML parsing problem.  I haven't used
HBaseTestingUtility (we start up a test cluster by extending
HBaseClusterTestCase), but I did notice something that looked strange in
your logs.  You can see that the namenode binds to 127.0.0.1:37523, and ZK
also uses 127.0.0.1, but then you see this:

10/10/06 20:53:14 INFO net.NetworkTopology: Adding a new node:
/default-rack/223.27.113.138:48680
10/10/06 20:53:14 INFO datanode.DataNode: New storage id
DS-517982013-127.0.1.1-48680-1286430794780 is assigned to data-node
223.27.113.138:48680

It looks like the datanode is binding to 223.27.113.138 instead of
127.0.0.1.  Then further down:

10/10/06 20:53:15 INFO hdfs.StateChange: BLOCK* NameSystem.allocateBlock:
/user/hudson/hbase.version. blk_6520143473758843527_1001
10/10/06 20:53:15 INFO hdfs.DFSClient: Exception in createBlockOutputStream
java.net.ConnectException: Connection refused
10/10/06 20:53:15 INFO hdfs.DFSClient: Abandoning block
blk_6520143473758843527_1001
10/10/06 20:53:15 INFO hdfs.DFSClient: Excluding datanode
223.27.113.138:48680
10/10/06 20:53:15 WARN namenode.FSNamesystem: Not able to place enough
replicas, still in need of 1

When my unit test cluster starts up everything binds to 127.0.0.1.  I'm not
sure if that's your problem, but it definitely looks like the DFS client
can't talk to the datanode for some reason.  Maybe someone who is more
familiar with HBaseTestingUtility will know what could cause this.  Do you
have hbase-default.xml and/or hbase-site.xml on your test classpath?  If so,
what do they contain?

-James



On Thu, Oct 7, 2010 at 1:59 AM, Imran M Yousuf <[email protected]> wrote:

> Just an update -
>
>...

Reply via email to