How to make a multi-historical node cluster?

Apache Druid Q&A Google Groups
#1

Hi,

I am new to druid and using the latest version (0.12.1).
I am done with single machine druid setup and multi-node deployment as given in the documentation Druid Cluster Tutorial

But the performance is not as good as expected and now I want to have multiple historical nodes but I cannot find any help (either in documentation or online forums) to do this.
I currently have 2 Historical nodes (historical+middlemanager as given in tutorial) and I can see that coordinator have registered both of the historical nodes (using the web portal).
But I can not divide loading data in nodes. Each time the same historical node do all the processing.

I know I am missing some parameters in configuration but I don’t know what to parameters I should change so that both the nodes works in parallel. Can someone help with an example of multi-historical nodes setup or point me to some helping source?

Regards

#2

Hi Junaid,

To add a new historical node you just start one up. If you see it in the coordinator web console then that means it’s being seen properly. The coordinator will balance data onto it gradually – you can accelerate this process by increasing maxSegmentsToMove (in the coordinator dynamic config, accessible via the pencil).