Deploy Kyuubi engines on YARN

Deploy Kyuubi engines on YARN#

Deploy Kyuubi Spark Engine on YARN#

Requirements#

To deploy Kyuubi’s Spark SQL engines on YARN, you’d better have cognition upon the following things.

  • Knowing the basics about Running Spark on YARN

  • A binary distribution of Spark which is built with YARN support

  • An active Apache Hadoop YARN cluster

  • An active Apache Hadoop HDFS cluster

  • Setup Hadoop client configurations at the machine the Kyuubi server locates

Configurations#

Environment#

Either HADOOP_CONF_DIR or YARN_CONF_DIR is configured and points to the Hadoop client configurations directory, usually, $HADOOP_HOME/etc/hadoop.

If the HADOOP_CONF_DIR points the YARN and HDFS cluster correctly, you should be able to run the SparkPi example on YARN.

$ HADOOP_CONF_DIR=/path/to/hadoop/conf $SPARK_HOME/bin/spark-submit \
    --class org.apache.spark.examples.SparkPi \
    --master yarn \
    --queue thequeue \
    $SPARK_HOME/examples/jars/spark-examples*.jar \
    10

If the SparkPi passes, configure it in $KYUUBI_HOME/conf/kyuubi-env.sh or $SPARK_HOME/conf/spark-env.sh, e.g.

$ echo "export HADOOP_CONF_DIR=/path/to/hadoop/conf" >> $KYUUBI_HOME/conf/kyuubi-env.sh

Spark Properties#

These properties are defined by Spark and Kyuubi will pass them to spark-submit to create Spark applications.

Note: None of these would take effect if the application for a particular user already exists.

  • Specify it in the JDBC connection URL, e.g. jdbc:kyuubi://localhost:10009/;#spark.master=yarn;spark.yarn.queue=thequeue

  • Specify it in $KYUUBI_HOME/conf/kyuubi-defaults.conf

  • Specify it in $SPARK_HOME/conf/spark-defaults.conf

Note: The priority goes down from top to bottom.

Master#

Setting spark.master=yarn tells Kyuubi to submit Spark SQL engine applications to the YARN cluster manager.

Queue#

Set spark.yarn.queue=thequeue in the JDBC connection string to tell Kyuubi to use the QUEUE in the YARN cluster, otherwise, the QUEUE configured at Kyuubi server side will be used as default.

Sizing#

Pass the configurations below through the JDBC connection string to set how many instances of Spark executor will be used and how many cpus and memory will Spark driver, ApplicationMaster and each executor take.

Name Default Meaning
spark.executor.instances 1 The number of executors for static allocation
spark.executor.cores 1 The number of cores to use on each executor
spark.yarn.am.memory 512m Amount of memory to use for the YARN Application Master in client mode
spark.yarn.am.memoryOverhead amMemory * 0.10, with minimum of 384 Amount of non-heap memory to be allocated per am process in client mode
spark.driver.memory 1g Amount of memory to use for the driver process
spark.driver.memoryOverhead driverMemory * 0.10, with minimum of 384 Amount of non-heap memory to be allocated per driver process in cluster mode
spark.executor.memory 1g Amount of memory to use for the executor process
spark.executor.memoryOverhead executorMemory * 0.10, with minimum of 384 Amount of additional memory to be allocated per executor process. This is memory that accounts for things like VM overheads, interned strings other native overheads, etc

It is recommended to use Dynamic Allocation with Kyuubi, since the SQL engine will be long-running for a period, execute user’s queries from clients periodically, and the demand for computing resources is not the same for those queries. It is better for Spark to release some executors when either the query is lightweight, or the SQL engine is being idled.

Tuning#

You can specify spark.yarn.archive or spark.yarn.jars to point to a world-readable location that contains Spark jars on HDFS, which allows YARN to cache it on nodes so that it doesn’t need to be distributed each time an application runs.

Others#

Please refer to Spark properties to check other acceptable configs.

Kerberos#

Kyuubi currently does not support Spark’s YARN-specific Kerberos Configuration, so spark.kerberos.keytab and spark.kerberos.principal should not use now.

Instead, you can schedule a periodically kinit process via crontab task on the local machine that hosts Kyuubi server or simply use Kyuubi Kinit.

Deploy Kyuubi Hive Engine on YARN#

Requirements#

To deploy Kyuubi’s Hive SQL engines on YARN, you’d better have cognition upon the following things.

Configurations#

Environment#

Either HADOOP_CONF_DIR or YARN_CONF_DIR is configured and points to the Hadoop client configurations directory, usually, $HADOOP_HOME/etc/hadoop.

If the HADOOP_CONF_DIR points to the YARN and HDFS cluster correctly, you should be able to run the Hive SQL example on YARN.

$ $HIVE_HOME/bin/hiveserver2
# In another terminal
$ $HIVE_HOME/bin/beeline -u 'jdbc:hive2://localhost:10000/default'
0: jdbc:hive2://localhost:10000/default> CREATE TABLE pokes (foo INT, bar STRING);
0: jdbc:hive2://localhost:10000/default> INSERT INTO TABLE pokes VALUES (1, 'hello');

If the Hive SQL passes and there is a job in YARN Web UI, it indicates the hive environment is good.

Required Environment Variable#

The HIVE_HADOOP_CLASSPATH is required, too. It should contain commons-collections-*.jar, hadoop-client-runtime-*.jar, hadoop-client-api-*.jar and htrace-core4-*.jar. All four jars are in the HADOOP_HOME.

For example, in Hadoop 3.1.0 version, the following is their location.

  • ${HADOOP_HOME}/share/hadoop/common/lib/commons-collections-3.2.2.jar

  • ${HADOOP_HOME}/share/hadoop/client/hadoop-client-runtime-3.1.0.jar

  • ${HADOOP_HOME}/share/hadoop/client/hadoop-client-api-3.1.0.jar

  • ${HADOOP_HOME}/share/hadoop/common/lib/htrace-core4-4.1.0-incubating.jar

Configure them in $KYUUBI_HOME/conf/kyuubi-env.sh or $HIVE_HOME/conf/hive-env.sh, e.g.

$ echo "export HADOOP_CONF_DIR=/path/to/hadoop/conf" >> $KYUUBI_HOME/conf/kyuubi-env.sh
$ echo "export HIVE_HADOOP_CLASSPATH=${HADOOP_HOME}/share/hadoop/common/lib/commons-collections-3.2.2.jar:${HADOOP_HOME}/share/hadoop/client/hadoop-client-runtime-3.1.0.jar:${HADOOP_HOME}/share/hadoop/client/hadoop-client-api-3.1.0.jar:${HADOOP_HOME}/share/hadoop/common/lib/htrace-core4-4.1.0-incubating.jar" >> $KYUUBI_HOME/conf/kyuubi-env.sh