Github user afine commented on a diff in the pull request:
https://github.com/apache/zookeeper/pull/186#discussion_r105300714
--- Diff:
src/java/main/org/apache/zookeeper/server/NettyServerCnxnFactory.java ---
@@ -162,7 +162,7 @@ public void messageReceived(ChannelHandlerContext ctx,
MessageEvent e)
+ " from " + ctx.getChannel());
}
NettyServerCnxn cnxn =
(NettyServerCnxn)ctx.getAttachment();
- synchronized(cnxn) {
+ synchronized(cnxn.getRpcLock()) {
processMessage(e, cnxn);
--- End diff --
>> Now one thread can be in processMessage while another thread is getting
stats about the connection. Is that ok?
> I believe this is OK. We can receive two concurrent stat commands, but we
only process one of them at a time. I'm also not a Netty wizard, so I could be
wildly wrong :)
This should be OK since all fields printed by `synchronized void
dumpConnectionInfo(PrintWriter pwriter, boolean brief)` are only updated in
`synchronized` methods.
Another option would be to put a shared lock on the command's execution
(shared with any other 4LW that acquires a lock on a connection other than its
own, I THINK `cons` is the only other one)? I think this may impact 4LW
performance slightly but it would prevent the need to add another lock to
`NettyServerCnxn`.
I think I like your way better.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
