Use this procedure to install your Big Match Ambari service into an existing cluster.
Before you begin
Note: Big Match should always be installed after you have a working Hortonworks or IBM® Open Platform cluster. Do not include the Big Match service as part of a fresh cluster installation. Use these steps only after you have a working
cluster.
Make sure the following services are installed in your cluster before you begin.
- Apache Spark™
- OOZIE
- HDFS
- HBASE
- YARN
- MAPREDUCE2
- ZOOKEEPER
Note: The following procedure includes reference to the terms
"master" and "slave" in an IT context. IBM does not support
the use of these terms as they diminish the horror of the dehumanizing practice of slavery. They are
used here because they appear within the Ambari
UI.
Procedure
- If not already installed, add the EPEL Repository to every host in the cluster by running
the following commands on each node.
sudo wget http://download.fedoraproject.org/pub/epel/6/x86_64/epel-release-6-8.noarch.rpm; sudo rpm -ivh epel-release-6-8.noarch.rpm
Note: If using RHEL 7, use the following
command:
sudo wget http://download.fedoraproject.org/pub/epel/7/x86_64/e/epel-release-7-8.noarch.rpm; sudo rpm -ivh epel-release-7-8.noarch.rpm
Note: Step 1 is required for the installer to work as it uses the python-requests package and
dependent libraries. The python-requests package and dependent libraries are available in the EPEL
repository.
-
Make sure that the Apache Spark Service and Spark Client are be installed before installing Big Match.
- Go to the Ambari UI and click Actions and then select
Add Service.
- On the Add Service Wizard, choose to install the service by checking
the box next to the Big Match service name. Click
Next.
-
IMPORTANT: On the Assign Masters screen, verify that the Big Match server is assigned to the same node where the Oozie server is installed. Click
Next.
- On the Assign Slaves and Clients screen, verify that the Big Match
client is installed on every node that runs one or more of the following components: HDFS_CLIENT,
DATANODE, HBASE_CLIENT, HBASE_MASTER, HBASE_REGIONSERVER, IBM_BIGMATCH_SERVER.
Note: In typical clusters, HBASE_REGIONSERVER is on its own node and is not designated as a client
node.
Because the HBase libraries must be added to that node, HBASE_REGIONSERVER must be a client node
for the Big Match client to be installed. Note that even though HBASE_REGIONSERVER becomes a client
node, the clients don't actually run on it and therefore does not affect memory usage.
- When you finish designating the appropriate nodes as clients, click
Next.
- On the Customize Services screen, select the Big
Match tab and then configure the properties for Advanced
bigmatch-config, Advanced bigmatch-install-config, and
Advanced pme-site. When all fields are filled, click
Next.
Note: Hovering over most fields shows a description of the property and the range of expected
values.
- If a configuration warning displays, and if the cluster is a production or performance
environment, follow the prompts to make the necessary changes before continuing. If the warnings are
related to memory or resources and this is a test environment where large resources are not
required, click Proceed Anyway. Refer to Hortonworks or IBM Open Platform documentation for details on handling any warnings.
- On the Review screen, verify the services that you are about to add
and click Deploy.
The installation begins. You can drill down to get status of the installation by
clicking the link under the Message header. Clicking the link shows log details
on the progress of the installation. This information can be used for debugging purposes if the
installation fails.
- When the installation completes, click Next.
- On the Summary screen, review the warning in red and make note of
the services that must be restarted. Click Complete.
- Restart the services notated with a yellow recycle icon before attempting to use Big Match.
Important: For Hortonworks installation, if the HBase Rest server is running before
installation of Big Match in a kerberized environment, then you must restart the server after
completing the installation. This ensures that the HBase Rest server picks up kerberos changes
provided in the Big Match installalation.