I just tried to test new flags, and as I understood, --ignore-sleeping should ignore time, when server sleeps for tick and this means if tick took 25 ms, sampler should not record this, if threshold is 50ms.

My commands:
/spark profiler --start --ignore-sleeping --only-ticks-over 50
/spark profiler --stop --order-by-time

Expected behaviour:
Empty logs.

Actual output:
https://spark.lucko.me/#yKgkSTTzBC
Command /spark tps shows that the maximum tick duration during a minute was 23.3ms, but logs show some data recorded.

Should be improved by bdc2641

@lucko
Data still recorded:

/spark profiler --ignore-sleeping --only-ticks-over 40
/spark profiler --stop
https://spark.lucko.me/#s13PlCV0f3

/spark profiler --ignore-sleeping --ignore-native --only-ticks-over 40
/spark profiler --stop
https://spark.lucko.me/#HVC2jVFZwv

But /tps command (by spark) shows no more than ~29 ms tick duration (max) after recording completed.

The same problem with /spark tickmonitor --without-gc --threshold-tick 40, it shows every tick, even if duration without sleeping < 50ms

Unfortunately, due to the way spark monitors ticks for this functionality, it is not possible to set a tick filter below 50ms (because the server sleeps to ensure all ticks are at least 50ms!)

Unfortunately, due to the way spark monitors ticks for this functionality, it is not possible to set a tick filter below 50ms (because the server sleeps to ensure all ticks are at least 50ms!)

But you can measure exact tick durations excluding sleep by events in recent versions, the same way you display them in /spark health (Tick Durations)