On 09/03/2021 13:12, Rémy Maucherat wrote:
On Tue, Mar 9, 2021 at 1:06 PM Mark Thomas <ma...@apache.org> wrote:
Hi Rory,
I have spent a lot of time of the last few days investigating some
networking issues with Tomcat and it looks like the root cause is in the
JRE. I have opened the following issue:
https://bugs.openjdk.java.net/browse/JDK-8263243
TL;DR it appears that ServerSocketChannel.accept() can return multiple,
valid SocketChannel instances for the same client connection. Start
trying to use those duplicate SocketChannel instances with a Selector
and stuff starts failing.
If the bug report is valid, it could be the root cause of all sorts of
strange networking bugs. I'd appreciate anything you can do to get the
right folks looking at the report. I am happy to provide any additional
information that may be required and/or test nay proposed fixes.
I don't get what the "wrk" test client is. Anyway, impressive stuff.
https://github.com/wg/wrk
I wonder if it happens with NIO2 as well.
Good question. The NIO2 acceptor is a little more complex but it should
be possible to tweak the test case to check.
Mark
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org
For additional commands, e-mail: dev-h...@tomcat.apache.org
