Cluster Deployment(Distributed)

This section describes how to deploy Trueno in a cluster environment, i.e. horizontal scaling.

Important notes:

• Trueno has 2 distributed sub-components that are resilient and distributed by design:
  • ElasticSearch
  • Spark
• When running Trueno in cluster mode, all machines will connect these sub-components and create your cluster.
• Any machine can receive and serve request, the backend storage will handle the persistence, consistency, sharding, and replication.

In order to deploy Trueno in a distributed way you will need follow these steps:

1. Install Trueno at each machine you want to deploy as part of the cluster.
2. Run the cluster setup coordinator, where the Spark Master will run.
3. Run other Trueno instances at other machines which will connect to the coordinator in order to setup the cluster.
4. (optional) Run a load balancer to redirect request to all machines in a balanced way.

Setup Guide

In this example there are 4 machines where you want to deploy your cluster, and you want to distribute the components between them.

NOTE: you can run all sub-components in all machines except the setup coordinator which must exist only in one machine:

machine	master	compute	backend	index
machine_1
machine_2
machine_3
machine_4

The previous table shows how the sub-components will be distributed by machine. To achieve this configuration, first a JSON configuration file is deployed in the cluster machines:

cluster.json

{
    "192.168.1.40": {"master": true, "compute": true},
    "192.168.1.41": {"compute": true,"backend": true},
    "192.168.1.42": {"compute": true,"backend": true,"index": true},
    "192.168.1.43": {"compute": true,"backend": true,"index": true}
}

After the configuration is deployed, start a Trueno instance on all machines.

NOTE The coordinator must be started first (in the JSON confguration example, 192.168.1.40).

NOTE: The machine name on the JSON file must match the hostname when you start the instance (using the -h flag).

Step 1

At machine 192.168.1.40:

  > trueno start -h 192.168.1.40 --cluster path/to/cluster.json

You will see the following line in the console output:

info: >> CLUSTER: Awaiting for cluster hosts ready signal

Step 2

At all other machines, start the Trueno instance the same way:

At machine 192.168.1.41:

  > trueno start -h 192.168.1.41 --cluster path/to/cluster.json

At machine 192.168.1.42:

  > trueno start -h 192.168.1.42 --cluster path/to/cluster.json

At machine 192.168.1.43:

  > trueno start -h 192.168.1.43 --cluster path/to/cluster.json

Now, at machine 192.168.1.40 (coordinator), you will see that the database will display a starting message. This means that the coordinator has connected all sub-components and is ready to receive requests.