노드 역할과 아키텍처

In an Elasticsearch cluster, different nodes can have specialized responsibilities. These are called node roles.

Assigning specific roles helps distribute workload, improve stability, and scale your cluster efficiently. It's like having specialists in a team!

A master node is responsible for managing the cluster state. It controls:

• Creating/deleting indices
• Tracking which nodes are part of the cluster
• Assigning shards to nodes

Without a stable master, your cluster can't function properly. It's crucial for cluster stability.

Data nodes are where your actual data lives. They perform the core search and indexing operations:

• Storing document shards
• Handling search requests
• Performing aggregations

If you need more storage or processing power for data, you add more data nodes.

Ingest nodes can preprocess documents before they are indexed. They use ingest pipelines to:

• Transform data (e.g., convert data types)
• Enrich data (e.g., add geo-location based on IP)
• Extract information (e.g., parse logs)

This offloads preprocessing from data nodes, improving their performance.

A coordinating node (or client node) acts as a smart load balancer. Any node can technically be a coordinating node, but dedicated ones are useful.

It handles:

• Receiving search requests from clients
• Distributing requests to data nodes
• Gathering results and sending them back to the client

They don't hold data or cluster state, focusing purely on request coordination.

In smaller clusters, a single node might perform multiple roles (e.g., master, data, ingest).

For larger, production-grade clusters, it's best practice to separate roles into dedicated nodes:

• Dedicated master nodes: Ensure cluster stability.
• Dedicated data nodes: Focus on storing and searching data.
• Dedicated ingest nodes: Handle data transformation.
• Dedicated coordinating nodes: Distribute client requests.

This improves performance, reliability, and scalability.

You define a node's roles in its elasticsearch.yml configuration file using the node.roles setting.

Here's a snippet for a data node:

For a small development or single-server setup, you might have one node performing all roles:

• Node 1: node.roles: [master, data, ingest]

This is simple but less resilient and scalable for production.

In a production environment, you'd separate concerns for better stability and scaling:

• 3 Master Nodes: node.roles: [master] (for fault tolerance)
• N Data Nodes: node.roles: [data] (scale based on data/query load)
• M Ingest Nodes: node.roles: [ingest] (scale based on ingestion needs)
• L Coordinating Nodes: node.roles: [] (optional, but good for client load balancing)

The node.roles: [] implies it's a coordinating-only node.

Which type of node is primarily responsible for managing the cluster state, such as creating indices and assigning shards?

Today, we learned about the different node roles in Elasticsearch:

• Master nodes manage cluster state.
• Data nodes store data and handle search/indexing.
• Ingest nodes preprocess documents.
• Coordinating nodes route requests.

Understanding these roles helps you design robust and scalable Elasticsearch clusters tailored to your needs. Separating roles is key for larger deployments!