This message was deleted.

jvm Config :

-server
-Xms4g
-Xmx4g
-XX:MaxDirectMemorySize=1280m
-Duser.timezone=UTC
-Dfile.encoding=UTF-8
-Djava.io.tmpdir=var/tmp
-Djava.util.logging.manager=org.apache.logging.log4j.jul.LogManager

How many segments do you have per historical ? And total ?

Cluster has 2 Historical Nodes. Historical Node which fails has 35,000 segments. Other Historical Node is same cluster has around 30,000 segment. [I don't see failure with this Historical Node].

druid.segmentCache.numBootstrapThreads
druid.segmentCache.numLoadingThreads

cool, what are the values set for this ?

I did not set those. Probably it will take default values.

you can try to set it up to higher value(It will take defaults wrt available cores) and increase the JVM size.

I've increase

druid.segmentCache.numLoadingThreads=4

and

druid.segmentCache.numBootstrapThreads

equivalent to 4 because it's default value is

druid.segmentCache.numLoadingThreads

] But It didn't help. I also change

balanceStrategy

to

random

from

cachingCost

Can you recommend If I need to tweak any parameters? I use m5.xlarge ec2 Instance as Historical Node. It has vcpu = 4 and Memory = 16 GB.

can you try with balanceStrategy default value cost ? Random will not help here much.