See all drill best practice FAQs.
Users who are running concurrent workloads on Drill frequently ask the Drill team this question.
This could be caused by an excessive number of threads running on a Drillbit. This can lead to excessive thread context switching. It results in the Drillbit not being able to send a heartbeat to ZooKeeper, leading ZooKeeper to believe the node has died. If you are running a concurrent workload, your queries may be using too many threads. You can confirm the excessive thread context switching by using system commands like sar and vmstat. If this is the case, you can tune the level of parallelism using the max_width_per_node and max_width_per_query options. See https://drill.apache.org/docs/configuring-resources-for-a-shared-drillbit/#configuring-parallelization for further details.
Retrieving data ...