Bernard Butler
2008-08-14 17:52:33 UTC
Hi,
I'm new to Hadoop - so hope you can help with this problem.
I'm trying to set up a small (2-zone) hadoop cluster on Solaris.
start-dfs.sh runs without error, e.g., it prints the following to the
screen:
master: starting datanode, logging to ..
slave: starting datanode, logging to ..
All looks well. However, when I check the DFS admin web page on the
master (on port 50070) it says
------------
Cluster Summary
1 files and directories, 0 blocks = 1 total. Heap Size is 18.5 MB / 448 MB
(4%)
Capacity : 0 KB
DFS Remaining : 0 KB
DFS Used : 0 KB
DFS Used% : 0 %
Live Nodes : 0
Dead Nodes : 0
There are no datanodes in the cluster
------------
I had a look in the datanode logs and they were empty on both master and
slave.
Running netstat -an on the master shows that it is listening on ports
50070, 50090 and 54310 (I changed fs.default.name to avoid a port
conflict). The slave has no hadoop-related ports active although there is
a single com.sun.management.jmxremote process running.
FYI a single-node pseudo-distributed installation worked fine on the
master. I'm running hadoop-0.17.1. I did not run start-mapred.sh.
Advice/suggestions would be very welcome.
Thanks,
B Butler
I'm new to Hadoop - so hope you can help with this problem.
I'm trying to set up a small (2-zone) hadoop cluster on Solaris.
start-dfs.sh runs without error, e.g., it prints the following to the
screen:
master: starting datanode, logging to ..
slave: starting datanode, logging to ..
All looks well. However, when I check the DFS admin web page on the
master (on port 50070) it says
------------
Cluster Summary
1 files and directories, 0 blocks = 1 total. Heap Size is 18.5 MB / 448 MB
(4%)
Capacity : 0 KB
DFS Remaining : 0 KB
DFS Used : 0 KB
DFS Used% : 0 %
Live Nodes : 0
Dead Nodes : 0
There are no datanodes in the cluster
------------
I had a look in the datanode logs and they were empty on both master and
slave.
Running netstat -an on the master shows that it is listening on ports
50070, 50090 and 54310 (I changed fs.default.name to avoid a port
conflict). The slave has no hadoop-related ports active although there is
a single com.sun.management.jmxremote process running.
FYI a single-node pseudo-distributed installation worked fine on the
master. I'm running hadoop-0.17.1. I did not run start-mapred.sh.
Advice/suggestions would be very welcome.
Thanks,
B Butler