On 31/07/2019 04:37, Kirill Ilyukhin wrote:
> Hello Mark,
>
> Please let me disagree with you. I have connected JVisualVM to Tomcat JVM,
> run the test with 10,000 connections, performed several explicit GCs and
> made a heap dump. All 10,000 WsSessions are in heap. They are referenced as
> follow
Hello Mark,
Please let me disagree with you. I have connected JVisualVM to Tomcat JVM,
run the test with 10,000 connections, performed several explicit GCs and
made a heap dump. All 10,000 WsSessions are in heap. They are referenced as
following:
this
<- wsSession - class:
org.apache.tomcat.w
On 30/07/2019 05:48, Kirill Ilyukhin wrote:
> Hello Mark,
>
> Please see the test case and Tomcat JVM heap dump screenshot attached.
> For sake of simplicity I do Thread.sleep() in client code instead of
> reading bytes from server.
That test case does not demonstrate a memory leak in either Tom
Hello Mark,
Please see the test case and Tomcat JVM heap dump screenshot attached.
For sake of simplicity I do Thread.sleep() in client code instead of
reading bytes from server.
Test configuration is the following:
Server version:Apache Tomcat/8.5.3
Server built: Jun 9 2016 11:1
On 26/07/2019 10:33, Kirill Ilyukhin wrote:
> Hello,
>
> When Tomcat receives WebSocket text message with invalid UTF-8, it closes
> this connection with NOT_CONSISTENT reason. But after that some objects
> (WsSession, UpgradeHandler, etc) stay in heap forever. They are referenced
> from AbstractP
Hello,
When Tomcat receives WebSocket text message with invalid UTF-8, it closes
this connection with NOT_CONSISTENT reason. But after that some objects
(WsSession, UpgradeHandler, etc) stay in heap forever. They are referenced
from AbstractProtocol's connections map.
This leak consistently happe