Cluster, node, and shard architecture in Elasticsearch - Step-by-Step Execution

Concept Flow - Cluster, node, and shard architecture

Start: Client sends data/query

↓

Cluster receives request

↓

Cluster routes to Node

↓

Node manages Shards

↓

Primary Shard

↓

Store data

↓

Node responds to Cluster

↓

Cluster sends response to Client

The cluster receives requests and routes them to nodes, which manage shards (primary and replicas) to store and retrieve data.

Execution Sample

Elasticsearch

Cluster -> Node -> Primary Shard -> Store Data
Cluster -> Node -> Replica Shard -> Store Copy

Shows how a cluster routes data storage to nodes and their shards.

Execution Table

Step	Component	Action	Result
1	Client	Sends data or query	Request sent to cluster
2	Cluster	Receives request	Routes to appropriate node
3	Node	Receives routed request	Determines shard to use
4	Primary Shard	Stores data or processes query	Data saved or query executed
5	Replica Shard	Stores copy of data	Backup created for fault tolerance
6	Node	Aggregates shard results	Prepares response
7	Cluster	Collects node responses	Forms final response
8	Client	Receives response	Operation complete

💡 Client receives response, ending the request cycle

Variable Tracker

Component	Start	After Step 2	After Step 4	After Step 5	Final
Cluster	Idle	Received request	Waiting for node	Waiting for node	Sent response
Node	Idle	Received request	Processed primary shard	Processed replica shard	Sent data to cluster
Primary Shard	Empty	N/A	Stored data	N/A	N/A
Replica Shard	Empty	N/A	N/A	Stored copy	N/A

Key Moments - 3 Insights

Why does the cluster route requests to nodes instead of handling data directly?

The cluster manages the overall system but nodes hold the actual data shards; routing to nodes allows distributed storage and processing, as shown in execution_table steps 2 and 3.

What is the difference between a primary shard and a replica shard?

Primary shards store the original data and handle write operations, while replica shards store copies for backup and fault tolerance, as seen in steps 4 and 5.

How does the system ensure data is not lost if a node fails?

Replica shards on other nodes keep copies of data, so if a node fails, the cluster can use replicas to maintain data availability, explained in step 5.

Visual Quiz - 3 Questions

Look at the execution_table, at which step does the primary shard store data?

A Step 4

B Step 3

C Step 5

D Step 2

💡 Hint

Check the 'Action' column for 'Stores data or processes query' in execution_table row 4

According to variable_tracker, what is the state of the replica shard after Step 5?

A Empty

B Processed primary shard

C Stored copy

D Sent data to cluster

💡 Hint

Look at the 'Replica Shard' row under 'After Step 5' in variable_tracker

If the cluster did not route requests to nodes, what would change in the execution_table?

A Step 1 would be repeated

B Steps 3 to 7 would be missing

C Step 8 would happen earlier

D Step 2 would be repeated

💡 Hint

Routing to nodes is shown in steps 3 to 7 in execution_table

Concept Snapshot

Cluster manages the whole system.
Nodes hold data shards.
Shards split data: primary (main) and replica (backup).
Cluster routes requests to nodes.
Nodes handle data storage and queries.
Replicas ensure data safety if nodes fail.

Full Transcript

In Elasticsearch, a cluster is a group of nodes working together. When a client sends data or a query, the cluster receives it and routes it to the right node. Each node manages shards, which are pieces of the data. Primary shards store the original data and handle writes. Replica shards keep copies for backup and fault tolerance. The node processes the request using these shards and sends the result back to the cluster, which then responds to the client. This system helps store data safely and handle queries efficiently.

Practice

1. What is a node in Elasticsearch cluster architecture?

easy

A. A single machine that is part of the cluster

B. A group of machines working together

C. A piece of data stored in the index

D. A backup copy of data

Solution

1. Step 1: Understand the definition of a node

   A node is one machine or server that joins the Elasticsearch cluster to share work.

2. Step 2: Differentiate node from cluster and shard

   A cluster is many nodes working together; shards are parts of data inside nodes.

3. Final Answer:

   A single machine that is part of the cluster -> Option A

4. Quick Check:

   Node = single machine [OK]

Hint: Node means one machine in the cluster [OK]

Common Mistakes:

• Confusing node with cluster
• Thinking node is a data piece
• Mixing node with shard

2. Which of the following is the correct way to describe a shard in Elasticsearch?

easy

A. A part of an index that stores data

B. A node in the cluster

C. A full copy of the entire index

D. A cluster of nodes

Solution

1. Step 1: Define what a shard is

   A shard is a smaller piece or partition of an index that holds part of the data.

2. Step 2: Compare with other options

   It is not a full copy (that's a replica), not a node or cluster.

3. Final Answer:

   A part of an index that stores data -> Option A

4. Quick Check:

   Shard = index part [OK]

Hint: Shard means a data part of an index [OK]

Common Mistakes:

• Thinking shard is a full index copy
• Confusing shard with node
• Mixing shard with cluster

3. Consider an Elasticsearch cluster with 3 nodes and an index configured with 5 primary shards and 1 replica per shard. How many total shards will be in the cluster?

medium

A. 5 shards

B. 10 shards

C. 6 shards

D. 15 shards

Solution

1. Step 1: Calculate total shards per primary and replica

   Each primary shard has 1 replica, so total shards = primary shards + replicas = 5 + 5 = 10 shards.

2. Step 2: Understand shard distribution across nodes

   Since there are 3 nodes, shards are distributed but total shards remain 10. However, the question asks total shards in cluster, including replicas.

3. Step 3: Re-check options

   15 shards says 15 shards, which is incorrect. The correct total is 10 shards.

4. Final Answer:

   10 shards -> Option B

5. Quick Check:

   Total shards = primary + replica = 5 + 5 = 10 [OK]

Hint: Total shards = primary shards x (1 + replicas) [OK]

Common Mistakes:

• Counting only primary shards
• Adding nodes to shard count incorrectly
• Confusing replicas with nodes

4. You have an Elasticsearch cluster with 2 nodes and an index with 4 primary shards and 1 replica. You try to assign 2 replicas per shard. What is the problem with this configuration?

medium

A. Replicas must be zero for cluster to work

B. Primary shards must be equal to replica shards

C. Number of nodes must be equal to number of shards

D. You cannot have more replicas than nodes minus one

Solution

1. Step 1: Understand replica shard limits

   Replicas cannot exceed the number of nodes minus one because replicas cannot be on the same node as primary.

2. Step 2: Apply to given cluster

   With 2 nodes, max replicas per shard = 1 (2 - 1). Trying 2 replicas is invalid.

3. Final Answer:

   You cannot have more replicas than nodes minus one -> Option D

4. Quick Check:

   Replicas ≤ nodes - 1 [OK]

Hint: Replicas max = nodes minus one [OK]

Common Mistakes:

• Setting replicas more than nodes - 1
• Confusing primary and replica counts
• Assuming replicas can be zero always

5. You want to design an Elasticsearch cluster for high availability with 4 nodes. You plan to create an index with 6 primary shards and 2 replicas per shard. What is the best shard and replica configuration to avoid shard allocation errors?

hard

A. 6 primary shards and 3 replicas per shard, total 24 shards

B. 6 primary shards and 2 replicas per shard, total 18 shards

C. 6 primary shards and 1 replica per shard, total 12 shards

D. 4 primary shards and 2 replicas per shard, total 12 shards

Solution

1. Step 1: Check replica limit based on nodes

   With 4 nodes, max replicas per shard = nodes - 1 = 3. So 2 replicas is allowed.

2. Step 2: Consider shard allocation rules

   Each shard copy must be on a different node. 6 primary shards + 2 replicas = 18 shards total, but 18 shards on 4 nodes means some nodes hold multiple shards.

3. Step 3: Avoid allocation errors by balancing shards

   Too many shards per node can cause performance issues. 1 replica (total 12 shards) is safer and avoids errors.

4. Final Answer:

   6 primary shards and 1 replica per shard, total 12 shards -> Option C

5. Quick Check:

   Replicas ≤ nodes - 1 and balanced shards [OK]

Hint: Replicas max = nodes - 1; balance shards per node [OK]

Common Mistakes:

• Setting replicas more than nodes - 1
• Ignoring shard distribution limits
• Choosing fewer shards than needed