9.1. Installing GeoMesa HBase

GeoMesa supports traditional HBase installations as well as HBase running on Amazon’s EMR , Hortonworks’ Data Platform (HDP), and the Cloudera Distribution of Hadoop (CDH). For details on bootstrapping an EMR cluster, see Bootstrapping GeoMesa HBase on AWS S3. For details on deploying to Cloudera CDH, see Deploying GeoMesa HBase on Cloudera CDH 5.X.

9.1.1. Installing the Binary Distribution

GeoMesa HBase artifacts are available for download or can be built from source. The easiest way to get started is to download the most recent binary version (2.0.2) from GitHub.

Extract it somewhere convenient:

# download and unpackage the most recent distribution:
$ wget "https://github.com/locationtech/geomesa/releases/download/geomesa_2.11-$VERSION/geomesa-hbase-dist_2.11-$VERSION-bin.tar.gz"
$ tar xvf geomesa-hbase-dist_2.11-$VERSION-bin.tar.gz
$ cd geomesa-hbase-dist_2.11-$VERSION
$ ls
bin/  conf/  dist/  docs/  examples/  lib/  LICENSE.txt  logs/

9.1.2. Building from Source

GeoMesa HBase may also be built from source. For more information refer to Building from Source in the developer manual, or to the README.md file in the the source distribution. The remainder of the instructions in this chapter assume the use of the binary GeoMesa HBase distribution. If you have built from source, the distribution is created in the target directory of geomesa-hbase/geomesa-hbase-dist.

More information about developing with GeoMesa may be found in the Developer Manual.

9.1.3. Configuration and Classpaths

GeoMesa HBase requires Hadoop and HBase jars and configuration files to be available on the classpath. This includes files such as the hbase-site.xml and core-site.xml files in addition to standard jars and libraries. Configuring the classpath is important if you plan to use the GeoMesa HBase command line tools to ingest and manage GeoMesa.

By default, GeoMesa HBase will attempt to read various HBase and Hadoop related environmental variables in order to build the classpath. You can configure environment variables and classpath settings in geomesa-hbase_2.11-$VERSION/conf/geomesa-env.sh or in your external env (e.g. bashrc file). The logic GeoMesa uses to determine which external entries to include on the classpath is:

1. If the environmental variables GEOMESA_HADOOP_CLASSPATH and GEOMESA_HBASE_CLASSPATH are set then GeoMesa HBase will use these variables to set the classpath and skip all other logic.

2. Next, if $HBASE_HOME and $HADOOP_HOME are set then GeoMesa HBase will attempt to build the classpath by searching for jar files and configuration in standard locations. Note that this is very specific to the installation or distribution of Hadoop you are using and may not be reliable.

3. If no environmental variables are set but the hbase and hadoop commands are available then GeoMesa will interrogate them for their classpaths by running the hadoop classpath and hbase classpath commands. This method of classpath determination is slow due to the fact that the hbase classpath command forks a new JVM. It is therefore recommended that you set manually set these variables in your environment or the conf/geomesa-env.sh file.

In addition, geomesa-hbase will pull any additional entries from the GEOMESA_EXTRA_CLASSPATHS environment variable.

Note that the GEOMESA_EXTRA_CLASSPATHS, GEOMESA_HADOOP_CLASSPATH, and GEOMESA_HBASE_CLASSPATH variables all follow standard Java Classpath conventions, which generally means that entries must be directories, JAR, or zip files. Individual XML files will be ignored. For example, to add a hbase-site.xml or core-site.xml file to the classpath you must either include a directory on the classpath or add the file to a zip or JAR archive to be included on the classpath.

Use the geomesa classpath command in order to see what JARs are being used.

A few suggested configurations are below:

When using EMR to install HBase or Hadoop there are AWS specific jars that need to be used (e.g. EMR FS). It is recommended to use EMR to install Hadoop and/or HBase in order to properly configure and install these dependencies (especially when using HBase on S3).

If you used EMR to install Hadoop and HBase, you can view their classpaths using the hadoop classpath and hbase classpath commands to build an appropriate classpath to include jars and configuration files for GeoMesa HBase:

export GEOMESA_HADOOP_CLASSPATH=$(hadoop classpath)
export GEOMESA_HBASE_CLASSPATH=$(hbase classpath)
export GEOMESA_HBASE_HOME=/opt/geomesa
export PATH="${PATH}:${GEOMESA_HBASE_HOME}/bin"

Configure GeoMesa to use pre-installed HBase and Hadoop distributions:

export HADOOP_HOME=/path/to/hadoop
export HBASE_HOME=/path/to/hbase
export GEOMESA_HBASE_HOME=/opt/geomesa
export PATH="${PATH}:${GEOMESA_HOME}/bin"

Configure the environment to use an HDP install

export HADOOP_HOME=/usr/hdp/current/hadoop-client/
export HBASE_HOME=/usr/hdp/current/hbase-client/
export GEOMESA_HBASE_HOME=/opt/geomesa
export PATH="${PATH}:${GEOMESA_HBASE_HOME}/bin"

If no HBase or Hadoop distribution is installed, try manually installing the JARs from maven:

export GEOMESA_HBASE_HOME=/opt/geomesa
export PATH="${PATH}:${GEOMESA_HBASE_HOME}/bin"
cd GEOMESA_HBASE_HOME
bin/install-hadoop.sh lib
bin/install-hbase.sh lib

You will also need to provide the hbase-site.xml file within a the GeoMesa conf directory, an external directory, zip, or JAR archive (an entry referencing the XML file directly will not work with the Java classpath).

When creating a zip or jar file, the hbase-site.xml should be at the root level of the archive and not nested within any packages or subfolders. For example:

$ jar tf my.jar
META-INF/
META-INF/MANIFEST.MF
hbase-site.xml
# try this
cp /path/to/hbase-site.xml ${GEOMESA_HBASE_HOME}/conf/

# or this
cd /path/to/hbase-conf-dir
jar cvf conf.jar hbase-site.xml
export GEOMESA_EXTRA_CLASSPATHS=/path/to/confdir:/path/to/conf.zip:/path/to/conf.jar

Due to licensing restrictions, dependencies for shape file support must be separately installed. Do this with the following commands:

$ bin/install-jai.sh
$ bin/install-jline.sh

9.1.4. Deploying the GeoMesa HBase Distributed Runtime JAR

GeoMesa uses an HBase custom filter to improve processing of CQL queries. In order to use the custom filter, you must deploy the distributed runtime jar to the HBase to the directory specified by the HBase configuration variable called hbase.dynamic.jars.dir. This is set to ${hbase.rootdir}/lib by default. Copy the distribute runtime jar to this directory as follows:

hadoop fs -put ${GEOMESA_HBASE_HOME}/dist/hbase/geomesa-hbase-distributed-runtime-$VERSION.jar ${hbase.dynamic.jars.dir}/

If running on top of Amazon S3, you will need to use the aws s3 command line tool.

aws s3 cp ${GEOMESA_HBASE_HOME}/dist/hbase/geomesa-hbase-distributed-runtime-$VERSION.jar s3://${hbase.dynamic.jars.dir}/

If required, you may disable distributed processing by setting the system property geomesa.hbase.remote.filtering to false. Note that this may have an adverse effect on performance.

9.1.5. Register the Coprocessors

GeoMesa utilizes server side processing to accelerate some queries. In order to utilize these features, the GeoMesa coprocessor must be registered on all GeoMesa tables or registered site-wide and the geomesa-hbase-distributed-runtime code must be available on the classpath or at an HDFS url, depending on the registration method used.

There are a number of ways to register the coprocessors, which are detailed later.

The following ways to register coprocessors can be done anytime and constitute the ‘upgrade path’, however, they may require HBase or tables to be taken offline.

  • Register Site-Wide using the hbase-site.xml
  • Register Per-Table using the hbase shell

The following ways to register coprocessors must be done before the tables are created.

  • Classpath Auto-Registration
  • System Property or geomesa-site.xml
  • DataStore Param Registration

There are two ways to get the coprocessor code on the classpath.

  • Modify the hbase-env.sh file and provide the path to the geomesa-hbase-distributed-runtime JAR in the HBASE_CLASSPATH property. If this method is used, the geomesa-hbase-distributed-runtime JAR must be available at the given location on all master and region servers.
  • If registering the coprocessors on a per-table basis using the hbase shell, it is possible to provide the HDFS path to the geomesa-hbase-distributed-runtime JAR that was deployed in Deploying the GeoMesa HBase Distributed Runtime JAR.

The easiest method to register the coprocessors is to specify the coprocessors in the hbase-site.xml. To do this simply add the coprocessors classname to the hbase.coprocessor.user.region.classes key.

<configuration>
  <property>
    <name>hbase.coprocessor.user.region.classes</name>
    <value>org.locationtech.geomesa.hbase.coprocessor.GeoMesaCoprocessor</value>
  </property>
</configuration>

All new and existing non-system tables will have access to the GeoMesa Coprocessor.

If your hbase instance is used for more than GeoMesa table or would like to utilize HDFS to deploy the geomesa-hbase-distributed-runtime JAR or for some other reason do not wish to register the coprocessor site wide you may configure the coprocessor on a per-table basis. This can be done by utilizing the the hbase shell as shown below. When specifying a coprocessor, the coprocessor must be available on the HBase classpath on all of the master and region servers or you must provide the HDFS URL for the geomesa-hbase-distributed-runtime JAR that was deployed in Deploying the GeoMesa HBase Distributed Runtime JAR.

To run the hbase shell simply execute:

$ ${HBASE_HOME}/bin/hbase shell
HBase Shell; enter 'help<RETURN>' for list of supported commands.
Type "exit<RETURN>" to leave the HBase Shell
hbase(main):001:0>

To get a list of the current tables run:

hbase(main):001:0> list
TABLE
geomesa
geomesa_QuickStart_id
geomesa_QuickStart_z2
geomesa_QuickStart_z3
4 row(s) in 0.1380 seconds

You will need to install the coprocessor on all table indexes list. The geomesa table in this example is the metadata table and does not need the coprocessor installed.

We use the alter command to modify the configuration of the tables. The coprocessor parameter in the alter command may be modified to change the registration of the GeoMesa coprocessors.

'coprocessor'=>'HDFS_URL|org.locationtech.geomesa.hbase.coprocessor.GeoMesaCoprocessor|PRIORITY|'

The ‘value’ of the coprocessor parameter has four parts, separated by |, two of which, HDFS_URL and PRIORITY, are configurable depending on your environment.

  • To provide the HDFS URL of the geomesa-hbase-distributed-runtime JAR replace HDFS_URL in the coprocessor value with the HDFS URL. This is only need if the geomesa-hbase-distributed-runtime JAR will not be on the classpath by other means.
  • To alter the priority (execution order) of the coprocessor change PRIRORITY to the desired value, this is optional and should be left blank if now used.
hbase(main):040:0> alter 'geomesa_QuickStart_id', METHOD => 'table_att', 'coprocessor'=>'|org.locationtech.geomesa.hbase.coprocessor.GeoMesaCoprocessor||'
Updating all regions with the new schema...
22/22 regions updated.
Done.
0 row(s) in 5.0000 seconds

hbase(main):041:0> alter 'geomesa_QuickStart_z2', METHOD => 'table_att', 'coprocessor'=>'|org.locationtech.geomesa.hbase.coprocessor.GeoMesaCoprocessor||'
Updating all regions with the new schema...
4/4 regions updated.
Done.
0 row(s) in 2.8850 seconds

hbase(main):042:0> alter 'geomesa_QuickStart_z3', METHOD => 'table_att', 'coprocessor'=>'|org.locationtech.geomesa.hbase.coprocessor.GeoMesaCoprocessor||'
Updating all regions with the new schema...
4/4 regions updated.
Done.
0 row(s) in 2.9150 seconds

To verify this worked successfully, run:

hbase(main):002:0> describe 'TABLE_NAME'
Table TABLE_NAME is ENABLED
TABLE_NAME, {TABLE_ATTRIBUTES => {coprocessor$1 => '|org.locationtech.geomesa.hbase.coprocessor.GeoMesaCoprocessor||'}
COLUMN FAMILIES DESCRIPTION
{NAME => 'm', BLOOMFILTER => 'ROW', VERSIONS => '1', IN_MEMORY => 'false', KEEP_DELETED_CELLS => 'FALSE', DATA_BLOCK_EN
CODING => 'NONE', TTL => 'FOREVER', COMPRESSION => 'NONE', MIN_VERSIONS => '0', BLOCKCACHE => 'true', BLOCKSIZE => '655
36', REPLICATION_SCOPE => '0'}
1 row(s) in 0.1940 seconds
If the geomesa-hbase-distributed-runtime JAR is available on the HBase classpath when the table is created then the GeoMesa coprocessors will be automatically registered for that table.

System Property or geomesa-site.xml are essentially the same as they utilize the same mechanism, but two different approaches.

If the Java system property geomesa.hbase.coprocessor.path is set in the environment running the GeoMesa ingest then the HDFS or S3 URL provided as the value will be automatically registered in the table descriptor. There are three to do this.

  • Set the system property in your shell environment using the JAVA_TOOL_OPTIONS environment variable.
export JAVA_TOOL_OPTIONS="${JAVA_TOOL_OPTIONS} -Dgeomesa.hbase.coprocessor.path=hdfs://path/to/geomesa-runtime.jar"
  • Set the system property in the geomesa-env.sh script.
setvar CUSTOM_JAVA_OPTS "${JAVA_OPTS} -Dgeomesa.hbase.coprocessor.path=hdfs://path/to/geomesa-runtime.jar"
  • Set the system property using the geomesa-site.xml configuration file.
<property>
    <name>geomesa.hbase.coprocessor.path</name>
    <value>hdfs://path/to/geomesa-runtime.jar</value>
    <description>HDFS or local path to GeoMesa-HBase Coprocessor JAR. If a local path is provided it must be the same for
        all region server. A path provided through the DataStore parameters will always override this property.
    </description>
    <final>false</final>
</property>
If you are using GeoMesa-HBase programmatically you can use the datastore parameter coprocessor.url to set an HDFS or S3 path to the geomesa-hbase-distributed-runtime JAR.

For more information on managing coprocessors see Coprocessor Introduction on Apache’s Blog.

9.1.6. Setting up the HBase Command Line Tools

Warning

To use HBase with the command line tools, you need to install the coprocessors first, as described above.

GeoMesa comes with a set of command line tools for managing HBase features located in geomesa-hbase_2.11-$VERSION/bin/ of the binary distribution.

Note

You can configure environment variables and classpath settings in geomesa-hbase_2.11-$VERSION/bin/geomesa-env.sh.

If desired, you may use the included script bin/geomesa-hbase configure to help set up the environment variables used by the tools. Otherwise, you may invoke the geomesa-hbase script using the fully-qualified path, and use the default configuration.

The tools will read the $HBASE_HOME and $HADOOP_HOME environment variables to load the appropriate JAR files for HBase and Hadoop. If installing on a system without HBase and/or Hadoop, the install-hbase.sh and install-hadoop.sh scripts in the bin directory may be used to download the required HBase and Hadoop JARs into the lib directory. You should edit this script to match the versions used by your installation.

Note

See Logging Configuration for information about configuring the SLF4J implementation.

Note

GeoMesa provides the ability to provide additional jars on the classpath using the environmental variable $GEOMESA_EXTRA_CLASSPATHS. GeoMesa will prepend the contents of this environmental variable to the computed classpath giving it highest precedence in the classpath. Users can provide directories of jar files or individual files using a colon (:) as a delimiter. These entries will also be added the the map-reduce libjars variable. Use the geomesa-hbase classpath command to print the final classpath that will be used when executing geomesa commands.

The tools also need access to the hbase-site.xml for your cluster. If $HBASE_HOME is defined, it will pick it up from there. However, it may not be available for map/reduce jobs. To ensure it’s availability, add it at the root level of the geomesa-hbase-datastore JAR in the lib folder:

$ zip -r lib/geomesa-hbase-datastore_2.11-$VERSION.jar hbase-site.xml

Warning

Ensure that the hbase-site.xml is at the root (top) level of your JAR, otherwise it will not be picked up.

Due to licensing restrictions, certain dependencies for shape file support must be separately installed. Do this with the following commands:

$ bin/install-jai.sh
$ bin/install-jline.sh

Test the command that invokes the GeoMesa Tools:

$ bin/geomesa-hbase
INFO  Usage: geomesa-hbase [command] [command options]
  Commands:
  ...

For more details, see HBase Command-Line Tools.

9.1.7. Installing GeoMesa HBase in GeoServer

The HBase GeoServer plugin is bundled by default in a GeoMesa binary distribution. To install, extract $GEOMESA_HBASE_HOME/dist/gs-plugins/geomesa-hbase-gs-plugin_2.11-$VERSION-install.tar.gz into GeoServer’s WEB-INF/lib directory. Note that this plugin contains a shaded JAR with HBase 1.2.3 bundled. If you require a different version, modify the pom.xml and build the GeoMesa HBase plugin project from scratch with Maven.

This distribution does not include the Hadoop or Zookeeper JARs; the following JARs should be copied from the lib directory of your HBase or Hadoop installations into GeoServer’s WEB-INF/lib directory:

(Note the versions may vary depending on your installation.)

  • hadoop-annotations-2.7.4.jar
  • hadoop-auth-2.7.4.jar
  • hadoop-common-2.7.4.jar
  • hadoop-mapreduce-client-core-2.7.4.jar
  • hadoop-yarn-api-2.7.4.jar
  • hadoop-yarn-common-2.7.4.jar
  • htrace-core-3.1.0-incubating.jar
  • commons-cli-1.2.jar
  • commons-io-2.5.jar (you may need to remove an older version (2.1) from geoserver)
  • hbase-common-1.2.6.jar
  • hbase-client-1.2.6.jar
  • hbase-server-1.2.6.jar
  • hbase-protocol-1.2.6.jar
  • metrics-core-2.2.0.jar
  • netty-3.6.2.Final.jar
  • netty-all-4.0.41.Final.jar
  • zookeeper-3.4.10.jar
  • commons-configuration-1.6.jar

You can use the bundled $GEOMESA_HBASE_HOME/bin/install-hadoop.sh script to install these JARs.

  • hadoop-annotations.jar
  • hadoop-auth.jar
  • hadoop-common.jar
  • protobuf-java.jar
  • commons-io.jar
  • hbase-server-1.2.6.jar
  • zookeeper-3.4.10.jar
  • commons-configuration-1.6.jar

The HBase data store requires the configuration file hbase-site.xml to be on the classpath. This can be accomplished by placing the file in geoserver/WEB-INF/classes (you should make the directory if it doesn’t exist). Utilizing a symbolic link will be use full here so any changes are reflected in GeoServer.

ln -s /path/to/hbase-site.xml /path/to/geoserver/WEB-INF/classes/hbase-site.xml
ln -s /usr/hdp/current/hbase-client/hbase-site.xml /path/to/geoserver/WEB-INF/classes/hbase-site.xml

Restart GeoServer after the JARs are installed.

9.1.7.1. Jackson Version

Warning

Some GeoMesa functions (in particular Arrow conversion) requires jackson-core-2.6.x. Some versions of GeoServer ship with an older version, jackson-core-2.5.0.jar. After installing the GeoMesa GeoServer plugin, be sure to delete the older JAR from GeoServer’s WEB-INF/lib folder.

9.1.8. Connecting to External HBase Clusters Backed By S3

To use a EMR cluster to connect to an existing, external HBase Cluster first follow the above instructions to setup the new cluster and install GeoMesa.

The next step is to obtain the hbase-site.xml for the external HBase Cluster, copy to the new EMR cluster and copy it into ${GEOMESA_HBASE_HOME}/conf. At this point you may run the geomesa-hbase command line tools.

If you wish to execute SQL queries using Spark, you must first zip the hbase-site.xml file for the external cluster:

zip hbase-site.zip hbase-site.xml

Then copy the zip file to ${GEOMESA_HBASE_HOME}/conf then add the zipped configuration file to the Spark classpath:

export SPARK_JARS=file:///opt/geomesa/dist/spark/geomesa-hbase-spark-runtime_2.11-${VERSION}.jar,file:///opt/geomesa/conf/hbase-site.zip

Then start up the Spark shell:

spark-shell --jars $SPARK_JARS

9.1.9. Configuring HBase on Azure HDInsight

HDInsight generally creates HBASE_HOME in HDFS under the path /hbase. In order to make the GeoMesa coprocessors and filters available to the region servers, use the hadoop filesystem command to put the GeoMesa JAR there:

hadoop fs -mkdir /hbase/lib
hadoop fs -put geomesa-hbase-distributed-runtime-$VERSION.jar /hbase/lib/