2018-03-15 11:13 GMT+01:00 Ivan Kelly <iv...@apache.org>: > > What is the difference in Channel#write/ByteBuf pooling.....in Java 9 ? > Sounds like it could be an issue in netty itself. Java 9 removed a > bunch of stuff around Unsafe, which I'm pretty sure netty was using > for ByteBuf. Have you tried setting the pool debugging to paranoid? > > -Dio.netty.leakDetection.level=paranoid >
only with 'advanced' , sorry. I will try again with paranoid I have tried to attach my laptop to the same BK cluster and create a reproducer client, no results !! Bookies do not break! it seems that the error is only between the machines of that group (network) It seems something very weird, maybe a mix of message size/network settings/Java GC (pooled ByteBuf heap bufs are released on finalize() as far as I have understood from Netty logs) I hope that with 'paranoid' I will have some more stacktrace to have code references Thank you Enrico > > i tried running my attempted repro in java9, but no cigar. > > > -Ivan >
