https://bz.apache.org/bugzilla/show_bug.cgi?id=63916
--- Comment #12 from Remy Maucherat <r...@apache.org> --- By default on my Linux the socket send buffer size is 2304. For NIO to have the same "performance" and behavior running the test [as seen with the debug logs] as NIO2, the send buffer size needs to be exactly >= to 4737. Note: CI used to show this behavior but that's no longer the case, now NIO and NIO2 are equal and APR is quite a bit behind. https://ci.apache.org/projects/tomcat/tomcat9/logs/4721/TEST-org.apache.catalina.nonblocking.TestNonBlockingAPI.NIO.txt https://ci.apache.org/projects/tomcat/tomcat9/logs/4721/TEST-org.apache.catalina.nonblocking.TestNonBlockingAPI.NIO2.txt https://ci.apache.org/projects/tomcat/tomcat9/logs/4721/TEST-org.apache.catalina.nonblocking.TestNonBlockingAPI.APR.txt I have yet to notice any performance impact at all in a "real" test, NIO async with HTTP/2 is very fine performance wise compared to NIO2, same as sendfile, etc. Also at this point given the difference and most importantly that NIO2 is also affected, I would say FreeBSD's Java IO might be broken. AFAIK it is not a platform directly supported by OpenJDK. -- You are receiving this mail because: You are the assignee for the bug. --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org