Updated on 2022-12-14 GMT+08:00

Query Performance Deterioration

Symptom

The query performance fluctuates when the query is executed in different query periods.

Possible Causes

During data loading, the memory configured for each executor program instance may be insufficient, resulting in more Java GCs. When GC occurs, the query performance deteriorates.

Troubleshooting Method

On the Spark UI, the GC time of some executors is obviously higher than that of other executors, or all executors have high GC time.

Procedure

Log in to Manager and choose Cluster > Services > Spark2x. On the displayed page. click the Configurations tab and then All Configurations, search for spark.executor.memory in the search box, and set its to a larger value.

Reference

None