You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We suspect this is due to precision lost from converting nanoseconds to milliseconds. For Stage Level All Metrics, the results are in nanoseconds, and for Stage level aggregated task metrics, the results are in milliseconds. Most tasks in this stage has < 1,000,000 nanoseconds = 1 ms shuffle write time, which will rounds to 0 ms.
Steps/Code to reproduce bug spark_rapids profiling -v -e <my_tools_repo>/core/src/test/resources/spark-events-profiling/rapids_join_eventlog.zstd
The text was updated successfully, but these errors were encountered:
Describe the bug
Profiler output shows inconsistent shuffle write write time results in
profiler.log
.Under
Stage Level All Metrics
, the total is 93 ms.Under
Stage level aggregated task metrics
, the total is 42ms.We suspect this is due to precision lost from converting nanoseconds to milliseconds. For
Stage Level All Metrics
, the results are in nanoseconds, and forStage level aggregated task metrics
, the results are in milliseconds. Most tasks in this stage has < 1,000,000 nanoseconds = 1 ms shuffle write time, which will rounds to 0 ms.Steps/Code to reproduce bug
spark_rapids profiling -v -e <my_tools_repo>/core/src/test/resources/spark-events-profiling/rapids_join_eventlog.zstd
The text was updated successfully, but these errors were encountered: