Indexing slow down when we increase data node


(Parthiban P) #1

We are using Elastic Search Nest client for our application , we did load testing testing on inserting documents into Elastic Search Cluster.

Whenever we increase the data node, performance get degraded.

Master Node : 1
Data Node : 1

Index Size 2Kb , Shard Count 4, replica 0

Test Case

5 parallel thread , each thread synchronously pushing records(10000)
We could able to insert 25 record per second .

When we add the one more data node to cluster and created new insert . We could able to insert only 4 records per second ..

Is there anything am i missing?


(Adrien Grand) #2

How long have you been running the test? Is 10000 your batch size?

The difference between your two setups is that in the first case, data is always indexed locally, while in the second case it sometimes need to go to the other node (but the reduction still looks big to me). So what I suspect is that the second setup may look like it cannot index as fast as the first setup due to this additionnal overhead, but on the other hand if you try to max out indexing speed by sending more data in parallel, then the second setup will perform better since it has more capacity overall (in particular 2x more computing power I assume).


(Parthiban P) #3

Yes that is the expectation .. But we couldn't attain the same performance with single data node . Configurations available both data nodes are same. What could be done to improve the performance?


(Xavier Facq) #4

Hi,

When I reindex a lot of documents I change this settings before the indexation:

  • Set number of replica to 0
  • Set the refresh interval to -1

Create bulk with:

  • refresh = false
  • ConsistencyLevel with WriteConsistencyLevel.ONE

Hope this help


(system) #5

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.