Druid cluster with 2 nodes, 1 Node with broker service and other node executing remaining 4 druid services (Coordinator, Overload, Historical, MiddleManager). EC2 machine type is t2.xlarge.
My target of ingestion data into Druid is 150 million records in 1 data source, to test Druid`s capability on consumption to response in Sub Seconds.
Problem
I have loaded 10M of records into Druids data source. After this benchmark, I am facing Java.io.IOException: No Space Left on device due to middle manager services. Can anyone guide me with suitable configurations to load data in this huge chunk.
I have tried following options as Hit & Try strategy:
- Increased JVM memory of middle manager & historical services
- Increased buffer size of historical node upto 300 GB
- Reboot EC2 machine and restarted running services
