Cloud-local is a collection of bash scripts to set up a single-node cloud on on your desktop, laptop, or NUC. Performance expectations are to be sufficient for testing things like map-reduce ingest, converters in real life with real files, and your own geoserver/iterator stack. This setup is preconfigured to run YARN so you can submit command line tools mapreduce jobs to it. Currently localhost ssh is NOT required so it will work on a NUC.
Cloud Local can be used to run single node versions of the following software:
- Hadoop HDFS
- YARN
- Accumulo
- HBase
- Spark (on yarn)
- Kafka
- GeoMesa
The main branch requires Hadoop 3.x and Accumulo 2.0 while the hadoop2 branch is based on Hadoop 2.x and Accumulo 1.x
A proxy server can be configured by using the standard http_proxy env var or the cloud-local specific cl_http_proxy env var. The cloud local specific variable takes precedence.
To prepare for the first run of cloud-local, you may need to unset environment variables HADOOP_HOME, ACCUMULO_HOME, ZOOKEEPER_HOME, HBASE_HOME, and others. If env |grep -i hadoop comes back empty, you should be good to go. You should also kill any instances of zookeeper or hadoop running locally. Find them with jps -lV.
When using this the first time...
git clone [email protected]:ccri/cloud-local.git
cd cloud-local
bin/cloud-local.sh init
By default only HDFS is started. Edit conf/cloud-local.conf to enable Accumulo, BHase, Kafka, GeoMesa, Spark and/or Zeppelin.
Cloud local sets a default accumulo instance name of "local" and password of "secret" which can be modified by editing the conf/cloud-local.conf file. If you want to change this you'll need to stop, clean, and reconfigure.
This init script does several things:
- configure HDFS configuration files
- format the HDFS namenode
- create a user homedir in hdfs
- initialize accumulo/hbase
- start up zookeeper, hadoop, and accumulo/hbase
- start kafka broker
- install and start Zeppelin
- install GeoMesa Accumulo iterators
- install GeoMesa command-line tools
After running init source the variables in your bashrc or other shell:
source bin/config.sh
Now you should have the environment vars set:
env | grep -i hadoop
Now you can run fun commands like:
hadoop fs -ls /
accumulo shell -u root
After installing it you should be able to reach your standard cloud urls:
- Accumulo: http://localhost:50095
- Hadoop DFS: http://localhost:50070
- Job Tracker: http://localhost:8088
- Zeppelin: http://localhost:5771
Options for using cloud-local.sh can be found by calling:
bin/cloud-local.sh help
You can also set CL_VERBOSE=1 env variable in conf/cloud-local.conf to increase messages
You can safely stop the cloud using:
bin/cloud-local.sh stop
You should stop the cloud before shutting down the machine or doing maintenance.
You can start the cloud back up using the analogous start option. Be sure that the cloud is not running (hit the cloud urls or ps aux|grep -i hadoop).
bin/cloud-local.sh start
If existing ports are bound to the ports needed for cloud-local an error message will be printed and the script will stop.
cloud-local allows you to modify the ports, hostname, and bind addresses in configuration or using variables in your env (bashrc). For example:
# sample .bashrc configuration
# offset all ports by 10000
export CL_PORT_OFFSET=10000
# change the bind address
export CL_BIND_ADDRESS=192.168.2.2
# change the hostname from localhost to something else
export CL_HOSTNAME=mydns.mycompany.com
Port offseting moves the entire port space by a given numerical amount in order to allow multiple cloud-local instances to run on a single machine (usually by different users). The bind address and hostname args allow you to reference cloud local from other machines.
WARNING - you should stop and clean cloud-local before changing any of these parameters since they will modify the config and may prevent cloud-local from cleanly shutting down. Changing port offsets is supported by XML comments in the accumulo and hadoop config files. Removing or changing these comments (CL_port_default) will likely cause failures.
If you have the environment variable GEOSERVER_HOME set you can use this parameter to start GeoServer at the same time but running in a child thread.
bin/cloud-local.sh start -gs
Similarly, you can instruct cloud-local to shutdown GeoServer with the cloud using:
bin/cloud-local.sh stop -gs
Additionally, if you need to restart GeoServer you may use the command regeoserver:
bin/cloud-local.sh regeoserver
The GeoServer PID is stored in $CLOUD_HOME/data/geoserver/pid/geoserver.pid and GeoServer's stdout is redirected to $CLOUD_HOME/data/geoserver/log/std.out.
Zeppelin is disabled by default.
Currently, we are using the Zeppelin distribution that includes all of the interpreters, and it is configured to run against Spark only in local mode. If you want to connect to another (real) cloud, you will have to configure that manually; see:
To enable GeoMesa's Spark-SQL within Zeppelin:
- point your browser to your local Zeppelin interpreter configuration
- scroll to the bottom where the Spark interpreter configuration appears
- click on the "edit" button next to the interpreter name (on the right-hand side of the UI)
- within the Dependencies section, add this one JAR (either as a full, local file name or as Maven GAV coordinates):
- geomesa-accumulo-spark-runtime_2.11-1.3.0.jar
- when prompted by the pop-up, click to restart the Spark intepreter
That's it! There is no need to restart any of the cloud-local services.
The cloud-local.sh script provides options for maintenance. Best to stop the cloud before performing any of these tasks. Pass in the parameter clean to remove software (but not the tar.gz's) and data. The parameter reconfigure will first clean then init.
When this git repo is updated, follow the steps below. The steps below will remove your data.
cd $CLOUD_HOME
bin/cloud-local.sh stop
bin/cloud-local.sh clean
git pull
bin/cloud-local.sh init
If you foobar your cloud, you can just delete everything and start over. You should do this once a week or so just for good measure.
cd $CLOUD_HOME
bin/cloud-local.sh stop #if cloud is running
rm * -rf
git pull
git reset --hard
bin/cloud-local.sh init
If you are using cloud-local within a virtual machine running you your local box you may want to set up port forwarding for port 50095 to see the accumulo monitor. For VirtualBox go to VM's Settings->Network->Port Forwarding section (name=accumulo, protocol=TCP, Host IP=127.0.0.1, Guest IP (leave blank), Guest Port=50095)