Cross-cluster search in Elasticsearch - Step-by-Step Execution

Concept Flow - Cross-cluster search

Start Search Request

↓

Identify Target Clusters

↓

Send Query to Local Cluster

↓

Send Query to Remote Clusters

↓

Collect Results from All Clusters

↓

Merge and Sort Results

↓

Return Combined Results to User

The search request is sent to local and remote clusters, results are gathered, merged, and returned as one combined response.

Execution Sample

Elasticsearch

GET /local_index,remote_cluster:remote_index/_search
{
  "query": { "match_all": {} }
}

This query searches both a local index and a remote cluster's index and returns combined results.

Execution Table

Step	Action	Target Cluster	Query Sent	Response Received	Result Status
1	Receive search request	Local cluster	match_all query on local_index	Pending	Waiting for responses
2	Send query to local cluster	Local cluster	match_all query on local_index	Results from local_index	Success
3	Send query to remote cluster	Remote cluster	match_all query on remote_index	Results from remote_index	Success
4	Merge results	Local + Remote	N/A	Combined sorted results	Success
5	Return results to user	Client	N/A	Combined results JSON	Complete

💡 All clusters responded successfully and results merged for final output.

Variable Tracker

Variable	Start	After Step 2	After Step 3	After Step 4	Final
local_results	empty	results from local_index	results from local_index	results from local_index	included in combined results
remote_results	empty	empty	results from remote_index	results from remote_index	included in combined results
combined_results	empty	empty	empty	merged local + remote	returned to user

Key Moments - 3 Insights

Why do we specify the remote cluster name before the index in the query?

Because the remote cluster name tells Elasticsearch where to send the query; see execution_table step 3 where the query targets 'remote_cluster:remote_index'.

What happens if the remote cluster does not respond?

The search waits for a timeout or error; results from the local cluster are returned alone. This is implied after step 3 if no response is received.

How are results from multiple clusters combined?

Results are merged and sorted by relevance or timestamp as shown in step 4 of the execution_table.

Visual Quiz - 3 Questions

Look at the execution_table, at which step are results from the remote cluster received?

A Step 3

B Step 2

C Step 4

D Step 5

💡 Hint

Check the 'Response Received' column for remote cluster results.

According to variable_tracker, what is the state of combined_results after Step 3?

A Contains remote results only

B Empty

C Contains local results only

D Contains merged local and remote results

💡 Hint

Look at the 'combined_results' row and the 'After Step 3' column.

If the remote cluster name is omitted in the query, what changes in the execution flow?

A Query fails with an error

B Query is sent to all clusters automatically

C Query is sent only to local cluster

D Query is sent only to remote cluster

💡 Hint

Refer to concept_flow where target clusters are identified explicitly.

Concept Snapshot

Cross-cluster search lets you query multiple Elasticsearch clusters at once.
Use the syntax 'remote_cluster:index' to target remote data.
Elasticsearch sends queries to all specified clusters.
Results are merged and sorted before returning.
This helps search across distributed data easily.

Full Transcript

Cross-cluster search in Elasticsearch allows a user to send a search query that targets both local and remote clusters. The process starts when the search request is received. The system identifies which clusters to query, including the local cluster and any remote clusters specified by name. The query is sent to the local cluster first, then to the remote clusters. Each cluster processes the query and returns results. These results are collected and merged, typically sorted by relevance or timestamp. Finally, the combined results are returned to the user as a single response. Variables like local_results and remote_results hold partial results before merging. The combined_results variable holds the final merged data. If a remote cluster does not respond, only local results are returned. The syntax for remote queries requires prefixing the index with the remote cluster name, such as 'remote_cluster:index'. This feature enables seamless searching across multiple Elasticsearch clusters from one query.