A query with multiple JOINs and aggregations runs slower than expected on a Photon cluster. After checking the Spark UI, some stages show 'photon=false'. What is the first diagnostic step?
-
A
Increase executor memory to force Photon to activate
-
B
Identify which specific operators triggered fallback by examining the physical query plan
-
C
Disable adaptive query execution (AQE) to prevent plan changes
-
D
Switch to a GPU-accelerated cluster instead