Understanding Clusters - padogrid/padogrid GitHub Wiki

◀️ Understanding Workspaces :link: Running Clusters ▶️

Under each workspace, you can create one or more clusters of your choice of the supported products. A cluster is created by running the create_cluster command which by default adds two (2) members. You can add more members by using the add_member command.

`create_cluster`

The create_cluster command creates a cluster for the specified product. If the -product option is not specified, then it creates a cluster for the product in the current cluster context. For example, if your current cluster context is Geode, then running create_cluster will create another Geode cluster. The supported products are coherence, geode, hazelcast, kafka, redis, spark, gemfire, hadoop, jet, mosquitto, and snappydata.

# Create a Geode cluster with the default name 'mygeode'.
create_cluster -product geode

# Switch cluster context and create another Geode cluster named 'mygeode2'
switch_cluster mygeode
create_cluster -cluster mygeode2

# Create a Hazelcast cluster by specifying the product
create_cluster -product hazelcast -cluster myhz2

Cluster Environment Variables

Interally, PadoGrid sets the following environment variables when you switch clusters. They determine the current cluster context.

Environment Variables	Descriptions	Default
PRODUCT	Product name. Valid values are `coherence`, `geode`, `hazelcast`, `kafka`, `redis`, `spark`, `gemfire`, `hadoop`, `jet`, `mosquitto`, and `snappydata`.	N/A
CLUSTER	Cluster name. If the `-cluster` is not specified then the PadoGrid commands default to this value.	Coherence: `mycoherence`, ComputeDB: `mysnappy`, Geode: `mygeode`, GemFire: `mygemfire`, Hazelcast Platform/IMDG: `myhz`, Hazelcast Jet: `myjet`, Mosquitto: `mymosquitto`, Redis: `myredis`, SnappyData: `mysnappy`, Spark: `myspark`
POD	Pod name. If the `-pod` option is not specified then the PadoGrid commands default to this value.	local

Cluster `setenv.sh`

Setting cluster specific environment variables such as JAVA_OPTS is done by adding them in <cluster-dir>/setenv.sh, which is sourced in by all PadoGrid commands after the workspace setenv.sh is sourced in. This means you can override workspace-level cluster settings from this file.

# Add or override cluster specific settings in the following file:
vi $PADOGRID_WORKSPACE/clusters/$CLUSTER/bin_sh/setenv.sh

# Or change directory and edit from there:
cd_cluster
vi bin_sh/setenv.sh

SnappyData `setenv.sh`

In addition to JAVA_OPTS, for SnappyData clusters, the following environment variables are available for overriding the default bootstrap options or including additional bootstrap options. You can set them in the cluster's setenv.sh file.

Variable	For	Description
LOCATOR_OPTS	Locators	An argument list of locator options.
LEAD_OPTS	Leaders	An argument list of leader options.
MEMBER_OPTS	Servers	An argument list of server (member) options.

Default Member Startup Classes

By default, PadoGrid runs each member using the same default product commands and server classes as shown below.

Coherence 14.x

"$JAVA" -server -showversion $JAVA_OPTS com.tangosol.net.DefaultCacheServer

Confluent 7.0.0+

KAFKA_OPTS=... kafka-server-start -daemon ...

Geode 1.x/GemFire 9.x

gfsh start locator $GFSH_OPTS $JAVA_OPTS
gfsh start server $GFSH_OPTS $JAVA_OPTS --classpath=$CLASSPATH

Hadoop 2.x, 3.x

# namenode
HADOOP_CONF_DIR=... HDFS_NAMENODE_OPTS=... hdfs --daemon start namenode
      
# datanode
HDFS_DATANODE_OPTS=... hdfs --config $HADOOP_CONF_DIR --daemon start datanode

Hazelcast 5.x

"$JAVA" -server $JAVA_OPTS com.hazelcast.core.server.HazelcastMemberStarter

Hazelcast IMDG 3.x

"$JAVA" -server $JAVA_OPTS com.hazelcast.core.server.StartServer

Hazelcast IMDG 4.x

"$JAVA" -server $JAVA_OPTS com.hazelcast.core.server.HazelcastMemberStarter

Hazelcast Jet 3.x

"$JAVA" -server $JAVA_OPTS com.hazelcast.jet.server.StartServer

Hazelcast Jet 4.x

"$JAVA" -server $JAVA_OPTS com.hazelcast.jet.server.JetMemberStarter

Kafka 2.8.0+

KAFKA_OPTS=... kafka-server-start.sh -daemon ...

Mosquitto 1.x, 2.x

mosquitto -d -p ... -c $CONFIG_FILE

Redis 6.x, 7.x

redis-server $CONFIG_FILE --port ...
redis-cli --cluster add-node ...

SnappyData 1.x

snappy locator start $SNAPPY_OPTS $JAVA_OPTS
snappy server start $SNAPPY_OPTS $JAVA_OPTS -classpath=$CLASSPATH
snappy leader start $SNAPPY_OPTS $JAVA_OPTS -classpath=$CLASSPATH

Spark 2.x, 3.x

SPARK_WORKER_OPTS=... start-daemon.sh start ...

where

GFSH_OPTS, HADOOP_CONF_DIR, HDFS_NAMENODE_OPTS, HDFS_DATANODE_OPTS, JAVA_OPTS, KAFKA_OPTS, SNAPPY_OPTS , and SPARK_WORKER_OPTS are bootstrap options provided in the workspace and cluster setenv.sh files. Note that there are additional Hadoop, Kafka, and Spark specific environment variables not shown in the above example.

Custom Member Startup Classes: `RUN_SCRIPT`

You can also use a custom class to start members by creating a script and setting the RUN_SCRIPT environment variable to the script's path in the setenv.sh file. You would normally place the script in the same cluster's bin_sh/ directory where the setenv.sh is located. The custom script inherits all the PadoGrid environment variables.

JAVA - Java executable
JAVA_OPTS - Java options
CLASSPATH - Java class paths

The script can use the above environment variables to start the application as shown in the example below. Note that the member automatically runs in the background.

#!/usr/bin/env bash

"$JAVA" -server $JAVA_OPTS com.myco.mycluster.MyServer

You would set the script path in the cluster setenv.sh file as follows:

# In setenv.sh
RUN_SCRIPT=$CLUSTER_DIR/bin_sh/your_script