Install CDH5 Accumulo 1.6 in CDH 5.1 QuickStart

[#1]

I am trying to install CDH5 Accumulo 1.6 in QuickStart 5.1 VM.  I don't have enough memory to run Cloudera Manager so I am doing the install manually.  I got the tar.gzip and exploded it in /opt.  i updated paths in accumulo-env.xml but am getting the 'can't load hadoop native librarry' error.  Accumulo will start and I can get a prompt, but when I try createtable it just hangs.

I was confused about how to use HADOOP_PREFIX with HADOOP_CONF_DR.  I ended up not setting HADOOP_PREFIX becuase of how it is used with HADOOP_CONF_DIR, which looks complete with HADDOP_PREFIX.

I did not see any steps for in the Accumulo install doc for working with the tar.gz.  Please let me know what the accumulo-env.xml settings should be when working with the QuickStart 5.1 VM.  

Thanks,

[#2]

Hi Don!

Which tarball did you download and attempt to use? Our latest release is available at http://archive.cloudera.com/accumulo-c5/cdh/5/accumulo-1.6.0-cdh5.1.0.tar.gz

Instead of manually updating paths in accumulo-site.xml or accumulo-env.sh, I recommend that you copy the example configuration files from $ACCUMULO_HOME/conf/examples/CDH5_packages to your configuration directory. There you can see the following values getting set:

HADOOP_PREFIX=/usr/lib/hadoop

HADOOP_CONF_DIR="$HADOOP_PREFIX/etc/hadoop"

After copying the files, I was able to init, start, and create a table via the shell.

An alternative option would be to use our RPM release directly, which already has all of these settings configured for you. It is available at http://archive.cloudera.com/accumulo-c5/redhat/6/x86_64/cdh/ and should work with the Quickstart VM without issue.

Let us know if you have any further questions, and good luck!

Mike

[#3]

Mike,

So I got the RMP and installed it.  I did a accumulo init and that seemed to work although I did have to run it as root.  However, when I tried to /usr/lib/accumulo/bin/start-all.sh I got a message that there were no tservers.  Please let me know what I am missing, or if there are steps for installing accumulo in a cdh5 Quick Start VM, please let me know.

Thanks,

Don Resnik

[#4]

Mike,

My mistake, I am getting the 'no tserver' error when I try to call accumulo shell -u root.  Which may make sense because it did not look like accumulo was started.  When I try to /usr/lib/accumulo/bin/start-all.sh I get an error about conf/master not being present.

Any help would be appreciated.

Thanks,

Don Resnik

[#5]

Did you install the accumulo-tserver, accumulo-master, etc.. RPMs as well?
It sounds like you did, but I just want to make sure.

At this point you should be able to use the service scripts to start each
role:
sudo /sbin/service accumulo-master start

Unfortunately, the bundled management scripts (start-here.sh, start-all.sh,
stop-all.sh) will not work without additional configuration specifying the
roles. Generally we recommend using Cloudera Manager to assign the roles.
If that is not possible then you can either manually start each service (as
above) or attempt to fill in proper values into
$ACCUMULO_CONF_DIR/{masters, slaves, monitor, tracer, gc}. You will also
need to make sure that the shell you are using to invoking the scripts is
pointed at ZOOKEEPER, HADOOP, and JAVA homes.

Mike

[#6]

Mike,

I did not install all the RMPs. I did not understand that all are required
even on a single node VM. I thought the main RPM contained everything I
needed and that other RMPs were for additional nodes with specific roles.

I will install all the RPMs and try again. Thanks for the detailed
response.

Don

[#7]

For clarification: The main RPM contains all of the binary files, the other
RPMs just contain service scripts.

[#8 · Accepted solution]

Mike,

I downloaded all the RPMs and used the service commands and accumulo started without error.

Thanks for your help.

Don