Setting up a coordinator node

The master nodes that we have seen previously are the most important for cluster stability. To prevent the queries and aggregations from creating instability in your cluster, coordinator (or client/proxy) nodes can be used to provide safe communication with the cluster.

Getting ready

You need a working Elasticsearch installation, as we described in the Downloading and installing Elasticsearch recipe in this chapter, and a simple text editor to change configuration files.

How to do it…

For the advance setup of a cluster, there are some parameters that must be configured to define different node types.

These parameters are in the config/elasticsearch.yml, file and they can be setup a coordinator node with the following steps:

1. Set up the node so that it's not a master, as follows:

node.master: false

2. Set up the node to not contain data, as follows:

node.data: false

How it works…

The coordinator node is a special node that works as a proxy/pass thought for the cluster. Its main advantages are as follows:

• It can easily be killed or removed from the cluster without causing any problems. It's not a master, so it doesn't participate in cluster functionalities and it doesn't contain data, so there are no data relocations/replications due to its failure.
• It prevents the instability of the cluster due to a developers' /users bad queries. Sometimes, a user executes aggregations that are too large (that is, date histograms with a range of some years and intervals of 10 seconds). Here, the Elasticsearch node could crash. (In its newest version, Elasticsearch has a structure called circuit breaker to prevent similar issues, but there are always borderline cases that can bring instability using scripting, for example. The coordinator node is not a master and its overload doesn't cause any problems for cluster stability.
• If the coordinator or client node is embedded in the application, there are less round trips for the data, speeding up the application.
• You can add them to balance the search and aggregation throughput without generating changes and data relocation in the cluster.