| Re: [jetty-users] Jetty 9.03 - seeing maximum websocket connection limit. why? |
|
Dear Users, It turns out that the “Port forwarding” service of the “NAT Network” feature in VirtualBox 5.1.26 (Running on Windows 10 LTSB) is the bottleneck. I ran the server on the physical Windows 10 machine itself and didn’t face the observed client-connect limit. Regards, Roland From: jetty-users-bounces@xxxxxxxxxxx [mailto:jetty-users-bounces@xxxxxxxxxxx]
On Behalf Of Joakim Erdfelt Jetty 9.0.3 is very old and represents the earliest implementations of the websocket stack. Please consider using something more recent. Especially since you are also using HTTPS (TLS/SSL) as you have an obligation to keep your Jetty and JVM up to date to maintain compatibility with the quickly evolving TLS/SSL landscape. Jetty 9.4.7.v20170914 was just released. Java 1.8u133 is odd, that doesn't look right (cannot find
a Java release with that update number) Release notes for 1.8u131 says it expired July 18, 2017 Release notes for 1.8u141 says it will expire next month on Oct 17, 2017 Also, since you stated you are running on Windows 10, keep in mind that many Windows network security tools / firewalls will get in your way and limit/prevent behaviors you might be attempting to rely on.
Joakim Erdfelt /
joakim@xxxxxxxxxxx On Tue, Sep 19, 2017 at 5:10 AM, Houtman, Roland <Roland.Houtman@xxxxxxxxxxxxxxx> wrote: Hi All, We’re running
1x Windows 10 inside virtualBox for the server instance
3x Windows 10 on a physical machine to run 25 client instances each Java 1.8u133 with 2GB ram. Jetty 9.03 We’re connecting software which opens per instance:
-
two websockets, and -
some servlets for listing folder content which is invoked regularly. And -
a servlet for downloading files. all via HTTPS When opening about 40 client instances,
Then, setting up the websocket fails more and more. With many tries I got to 52 client software instances running. The server software is still having 1 GB free heapspace. And no noticeable CPU usage Which machine the client software runs from(i.e. either single client instance running,or 25 instances running), has no impact on success rate. Does anyone know why this limit is surfacing? I already read somewhere to increase the QueuedThreadPool. I set it (from default 200) to 750, but there was no noticeable difference.
Kind regards, Roland
|