Increasing heap usage with recording stream

[Gunnar Morling]

Hi,

Here's a basic reproducer for the behaviour I observe:

cat > RecordingStreamTest.java <<EOF
import jdk.jfr.consumer.RecordingStream;

public class RecordingStreamTest {

public static byte[] bytes;

public static void main(String[] args) throws Exception {
RecordingStream stream = new RecordingStream();
stream.enable("jdk.ObjectAllocationInNewTLAB");
stream.enable("jdk.ObjectAllocationOutsideTLAB");
stream.onEvent(re -> {});
stream.startAsync();

for (int i = 1; i <= 100_000_000; i++) {
bytes = new byte[1024];
}

System.gc();

for (int i = 1; i <= 100_000_000; i++) {
bytes = new byte[1024];
}

stream.close();
stream.awaitTermination();

System.gc();
}
}
EOF

javac RecordingStreamTest.java

java -XX:+HeapDumpAfterFullGC
-XX:StartFlightRecording=duration=2m,name=MyRecording,filename=myrecording.jfr,settings=default
RecordingStreamTest

The heap dump after the second full GC contains 500K RecordedEvent
instances, and the flight recording shows that the heap size after GC is
constantly increasing. Is this a usage error on my side (if so, what should
I do to prevent this?), or is it actually a memory leak somewhere in the
recording stream code?

Many thanks for any help,

--Gunnar


Am Mo., 7. Dez. 2020 um 10:20 Uhr schrieb Gunnar Morling <
gunnar at hibernate.org>:
>
> Hi,
>
> I'm using a JDK 14 RecordingStream, and I observe an ever-increasing heap
usage as my application keeps running. The reference chain is
>
>     JFR Event Stream Thread -> jdk.jfr.internal.consumer.ChunkParser ->
parsers jdk.jfr.internal.LongMap -> objects java.lang.Object[512]  ->
jdk.jfr.internal.consumer.EventParser -> cached
jdk.jfr.consumer.RecordedEvent[131072]
>
> I.e. that array "RecordedEvent[] cached" is continuously growing. Is this
a usage error on my end? I was looking for some sort of flush() or similar
API on RecordingStream, but can't find such a method.
>
> Any help would be appreciated.
>
> Thanks a lot,
>
> --Gunnar