Hi, according to the guide https://druid.apache.org/docs/latest/tutorials/cluster#start-data-server, the historical and middleManager servers are started together using the bin/start-cluster-data-server script. However, we want to implement a separate start for the historical and middleManager servers on the same node, so that we can restart only the historical server without interrupting ingestion tasks on the middleManager. Is it possible to do this, or is it best practice to always start both the historical and middleManager servers together on the same node when they are acting as a data node, using the bin/start-cluster-data-server script?

Although the

start-cluster-data-server

script starts both processes, they are still independent. So you can stop and restart only the historical even if it was started with that script.