Design

Apache Mesos Framework support scheduling and running Apache Ignite nodes in a Mesos cluster. This develop a Mesos Scheduler, a Mesos Task.

Goal Scenario

User starts Ignite Mesos Framework, framework scheduler is started and registers in the Mesos Cluster. Once registration is successful the framework will start to launch tasks on nodes. The framework will begin expanding the cluster to the desired total number of nodes. If a node were to be lost, the framework would first attempt to restart the node and the framework will remove the lost node from the cluster.

Components

Mesos Framework

Multiple instances of the framework can be run for multiple topology Frameworks will re-register with Mesos. Fault Tolerant, to restart failed Apache Ignite node tasks.

Mesos Scheduler

The Mesos scheduler is the component with the most high-level intelligence in the framework. It will need to possess the ability to start Apache Ignite cluster and distribute the correct configuration and packages to all started nodes.

Responsibilities

• Restarting failed nodes
• Providing configuration to nodes (JVM OPTS)
• Auto scaling
• Providing package to slave nodes. (ignite package, users libs, config file)

Mesos Task 

The Mesos Task will be responsible for running container with required configuration.

Running Ignite Framework

For running Ignite Framework requires started Apache Mesos Cluster. Instruction how setting up a cluster: https://docs.mesosphere.com/getting-started/datacenter/install/.

Run the Framework via Marathon

Currently the recommended way to run the Ignite Framework is via Marathon. 

1. Install marathon. See https://docs.mesosphere.com/getting-started/datacenter/install/ marathon section.

2. Download or copy the json properties. Update the MESOS_MASTER_URL and any other parameters which would like to change. See Properties section.

   { 
     "id": "ignition", 
     "instances": 1, 
     "cpus": 2, 
     "mem": 2048, 
     "ports": [0], 
     "uris": [ 
       "https://s3.amazonaws.com/ignite-mesos-ip/ignite-mesos-1.1.0.jar" 
     ], 
     "env": { 
       "IGNITE_NODE_COUNT": "3", 
       "MESOS_MASTER_URL": "zk://localhost:2181/mesos", 
       "IGNITE_RUN_CPU_PER_NODE": "2", 
       "IGNITE_MEMORY_PER_NODE": "2048", 
       "IGNITE_USERS_LIBS_URL": "https://s3.amazonaws.com/ignite-mesos-ip/libs.zip",
       "IGNITE_CONFIG_XML_URL": "https://s3.amazonaws.com/ignite-mesos-ip/default-config.xml",
       "IGNITE_VERSION": "1.0.5"
     }, 
     "cmd": "java -jar ignite-mesos-1.1.0.jar" 
   } 

3. POST the config to Marathon by CURL or other tools. 

   curl -X POST -H "Content-type: application/json" --data-binary @marathon.json http://<marathon-ip>:8080/v2/apps/

   Make sure thath Ignite cluster is running, for this:

4. Open Marathon UI at http://<marathon-ip>:8080. Make sure that exists application with name "Ignition".

Run JAR file

1. Build ignite-mesos module.
2. Run IgniteFramework use the followin command.

       java -jar ignite-mesos-1.0.0-snapshot-jar-with-dependencies.jar 

    Make sure that Ignite cluster is running, for this:

    3. Open Mesos console at http://<master-ip>:5050. If all work correctly tasks with "Ignite node N" should have state "RUNNING" or "STAGING".

    4. Go to sandbox.

    5. Look at stdout and stderr logs.

Properties

The following properties can be used to configure the framework. The properties can be passed to how property of file or environment variables.

IGNITE_RUN_CPU_PER_NODE

UNLIMITED

2

IGNITE_MEMORY_PER_NODE

UNLIMITED

1024

IGNITE_DISK_SPACE_PER_NODE

1024

1024

IGNITE_NODE_COUNT

5

IGNITE_TOTAL_CPU

UNLIMITED

5

IGNITE_TOTAL_MEMORY

UNLIMITED

16024

IGNITE_TOTAL_DISK_SPACE

UNLIMITED

5024

IGNITE_MIN_CPU_PER_NODE

1

4

IGNITE_MIN_MEMORY_PER_NODE

256

1024

IGNITE_VERSION

latest

1.0.5

IGNITE_WORK_DIR

ignite-release

IGNITE_XML_CONFIG

/opt/ignite/ignite-config.xml

IGNITE_USERS_LIBS

/opt/libs/

MESOS_MASTER_URL

zk://localhost:2181/mesos

176.0.1.45:2181

Resources

1. JIRA
2. Apache Mesos
3. Setting up a Mesosphere Cluster