Memory Running Out During Automatic Tuning
Symptom
The system remains unresponsive for an extended period during automatic tuning of Spark or Hive. The dmesg log contains OOM information and process termination information.
Possible Causes
The parameter is set to a large value for automatic tuning. As a result, the required memory is much larger than the available system memory.
Troubleshooting Procedure
- Restart the suspended application.
- Modify the application parameter setting based on your actual environment conditions (using Spark as an example).
Search the application configuration file param_xxx.json for the parameters spark.executor.memory and spark.executor.instances. Set appropriate parameter values based on your environment specifications.
The spark.executor.memory parameter indicates the memory required by an application instance, and the spark.executor.instances parameter indicates the number of instances used by the application. The product of the two parameters is the total memory space consumed by the application.
- Restart the automatic tuning task.