8
votes

I'm struggling to setup a Hbase distributed cluster with 2 nodes, one is my machine and one is the VM, using the "host-only" Adapter in VirtualBox.

My problem is that the region server (from VM machine) can't connect to Hbase master running on host machine. Although in Hbase shell I can list, create table, ..., in regionserver on VM machine ('slave'), the log always show

org.apache.hadoop.hbase.regionserver.HRegionServer: Unable to connect to master.      Retrying. Error was:
java.net.ConnectException: Connection refused

Previously, I've successfully setup Hadoop, HDFS and MapReduce on this cluster with 2 nodes named as 'master', and 'slave', 'master' as master node and both 'master' and 'slave' work as slave nodes, these names bound to the vboxnet0 interface of VirtualBox (the hostnames in /etc/hostname is different). I'also specify the "slave.host.name" property for each node as 'master' and 'slave'.

It seems that the Hbase master on the 'master' always run with 'localhost' hostname, from the slave machine, I can't telnet to the hbase master with 'master' hostname. So is there any way to specify hostname use for Hbase master as 'master', I've tried specify some properties about DNS interface for ZooKeeper, Master, RegionServer to use internal interface between master and slave, but it still does not work at all.

/etc/hosts for both as something like

127.0.0.1   localhost
127.0.0.1   ubuntu.mymachine

# For Hadoop
192.168.56.1 master
192.168.56.101 slave

# The following lines are desirable for IPv6 capable hosts
::1     ip6-localhost ip6-loopback
fe00::0 ip6-localnet
ff00::0 ip6-mcastprefix
ff02::1 ip6-allnodes
ff02::2 ip6-allrouters
4

4 Answers

4
votes

The answer that @Infinity provided seems to belong to version ~0.9.4.

For version 1.1.4.

according to the source code from

org.apache.hadoop.hbase.master.HMaster

the configuration should be:

  <property>
    <name>hbase.master.hostname</name>
    <value>master.local</value>
    <!-- master.local is the DNS name in my network pointing to hbase master -->
  </property>

After setting this value, region servers are able to connect to hbase master; however, in my environment, the region server complained about:

com.google.protobuf.ServiceException: java.net.SocketException: Invalid argument

The problem disappeared after I installed oracle JDK 8 instead of open-jdk-7 in all of my nodes.

So in conclusion, here is my solution:

  • use dns name server instead of setting /etc/hosts, as hbase is very picky on hostname and seems requires DNS lookup as well as reverse DNS lookup.

  • upgrade jdk to oracle 8

  • use the setting item mentioned above.

2
votes

My host file is like

127.0.0.1 localhost

192.168.2.118 shashwat.machine.com shashwat

make your hosts file as following:

127.0.0.1 localhost

For Hadoop

192.168.56.1 master

192.168.56.101 slave

and in hbase conf put following entries :

<property>
<name>hbase.rootdir</name>
<value>hdfs://master:9000/hbase</value>
</property>
<property>
<name>hbase.master</name>
<value>master:60000</value>
<description>The host and port that the HBase master runs at.</description>
</property>
<property>
<name>hbase.regionserver.port</name>
<value>60020</value>
<description>The host and port that the HBase master runs at.</description>
<property>
<name>hbase.cluster.distributed</name>
<value>true</value>
</property>
<property>
<name>hbase.tmp.dir</name>
<value>/home/cluster/Hadoop/hbase-0.90.4/temp</value>
</property>
<property>
<name>hbase.zookeeper.quorum</name>
<value>master</value>
</property>
<property>
<name>dfs.replication</name>
<value>2</value>
</property>
<property>
<name>hbase.zookeeper.property.clientPort</name>
<value>2181</value>
<description>Property from ZooKeeper's config zoo.cfg.
The port at which the clients will connect.
</description>
</property>

If you are using localhost anywhere remove that and replace it with "master" which is name for namenode in your hostfile....

one morething you can do

sudo gedit /etc/hostname

this will open the hostname file bydefault ubuntu will be there so make it master. and restart your system.

For hbase specify in "regionserver" file inside conf dir put these entries

master slave

and restart.everything.

0
votes

There are two things that fix this class of problem for me:

1) Remove all "localhost" names, only have 127.0.0.1 pointing to the name of the hmaster node.

2) run "hostname X" on your hbase master node, to make sure the hostname matches what is in /etc/hosts.

Not being a networking expert, I can't say why this is important, but it is :)

0
votes

Most of the time the error is coming from Zookeeper that send a wrong hostname.

You can check what Zookeeper sends as HBase master host:

Find Zookeeper bin folder:

bin/zkCli.sh -server 127.0.0.1:2181
get /hbase/master

This should give you the HBase master IP that answer Zookeeper, so this IP must be accessible.