This is it: a Docker multi-container environment with Hadoop (HDFS), Spark and Hive. But without the large memory requirements of a Cloudera sandbox. (On my Windows 10 laptop (with WSL2) it seems to consume a mere 3 GB.)
To deploy an the HDFS-Spark-Hive cluster, run:
docker-compose up
Run example wordcount job:
make wordcount
Or deploy in swarm:
docker stack deploy -c docker-compose-v3.yml hadoop
docker-compose creates a docker network that can be found by running docker network list, e.g. dockerhadoop_default.
Run docker network inspect on the network (e.g. dockerhadoop_default) to find the IP the hadoop interfaces are published on. Access these interfaces with the following URLs:
- Namenode: http://<dockerhadoop_IP_address>:9870/dfshealth.html#tab-overview
- History server: http://<dockerhadoop_IP_address>:8188/applicationhistory
- Datanode: http://<dockerhadoop_IP_address>:9864/
- Nodemanager: http://<dockerhadoop_IP_address>:8042/node
- Resource manager: http://<dockerhadoop_IP_address>:8088/
The configuration parameters can be specified in the hadoop.env file or as environmental variables for specific services (e.g. namenode, datanode etc.):
CORE_CONF_fs_defaultFS=hdfs://namenode:8020
CORE_CONF corresponds to core-site.xml. fs_defaultFS=hdfs://namenode:8020 will be transformed into:
<property><name>fs.defaultFS</name><value>hdfs://namenode:8020</value></property>
To define dash inside a configuration parameter, use triple underscore, such as YARN_CONF_yarn_log___aggregation___enable=true (yarn-site.xml):
<property><name>yarn.log-aggregation-enable</name><value>true</value></property>
The available configurations are:
- /etc/hadoop/core-site.xml CORE_CONF
- /etc/hadoop/hdfs-site.xml HDFS_CONF
- /etc/hadoop/yarn-site.xml YARN_CONF
- /etc/hadoop/httpfs-site.xml HTTPFS_CONF
- /etc/hadoop/kms-site.xml KMS_CONF
- /etc/hadoop/mapred-site.xml MAPRED_CONF
If you need to extend some other configuration file, refer to base/entrypoint.sh bash script.