Hi Job, Is lookup setup correctly, such that the nodes can do forward and reverse lookup for the IPs in the cluster? Just put all nodes in /etc/hosts on all machines (quite doable for a 2 node cluster).
If you switch HBase (and Zookeeper) logging to DEBUG, you may see a bit more of what does and does not happen. Hope this helps. Friso On 17 dec 2010, at 11:24, Job wrote: > Hi, > > I have setup a simple 2 node cluster using cloudera's distribution CDh3 > of hadoop, hbase and zookeeper. > The system seems to be running fine. (http://192.168.10.139:60010 show > that the master is running with one regionserver, the 'zk dump' shows > the master address: "Master address: 192.168.10.139:60000" > > On the server if check the HBase shell status, it looks all fine. I can > telnet from my own host to the master's IP: 192.168.10.139:60000 > > But next, from my code (using the same versions of the libs) when I try > create a connection to this server: > > HBaseConfiguration config = new HBaseConfiguration(); > config.set("hbase.zookeeper.quorum", zookeeper); > admin = new HBaseAdmin(config); > > The log tells me, that the zookeeper session is being created (which > indeed shows up in de zk-dump and log on the server) but hangs on > 'getMaster': > > org.apache.hadoop.hbase.client.HConnectionManager$TableServers - > getMaster attempt 0 of 10 failed; retrying after sleep of 2000 > > etc. > > It never receives the actual connection, also it doesn't tell if it > could retrieve the master address, and if it tries to connect to the > correct server. > > Any idea what can be wrong here? - or pointers to where I should look > for debugging the problem. > > Thanks! > > Job >
