https://bz.apache.org/bugzilla/show_bug.cgi?id=63306

            Bug ID: 63306
           Summary: Memory leak during websocket connection close
           Product: Tomcat 7
           Version: 7.0.82
          Hardware: Other
                OS: Linux
            Status: NEW
          Severity: critical
          Priority: P2
         Component: WebSocket
          Assignee: dev@tomcat.apache.org
          Reporter: 05.shail...@gmail.com
  Target Milestone: ---

Created attachment 36506
  --> https://bz.apache.org/bugzilla/attachment.cgi?id=36506&action=edit
Heap dump details

We use tomcat as web application server and use websockets to send updates to
the client. Websocket clients are written in the .net and connect to
application in turn tomcat websockets.

Websocket clients sometimes on wireless network. We have seen that there are
few thousands (20-40k) connections and disconnections by around 1000 users
(websocket client sessions).

We observed that application memory in on increasing trend day by day and after
5 days we took heap dump and found that there is memory leak in tomcat.

Please find attached screenshots of heapdump opened in Eclipse Memory Analyzer.

Heapdumps are quite huge and hence I can not upload.

Also in the catalina.out I have seen following log indicating that the servelt
output stream could not be closed properly.

org.apache.tomcat.websocket.server.WsRemoteEndpointImplServer doClose
INFO: Failed to close the ServletOutputStream connection cleanly

However above log messages are comparatively far less than the SecureNioChannel
and NIOProcessor instance lingering in the Map.Entry.

For the SecureSocketChannel.open flag is set to false and underlying socket is
also closed as highlighted in the attachments, however Map entries were not
removed which has caused memory leak.

With this memory leak, we can not take risk to run tomcat continued for days
and hence need to restart tomcat after few days.

We can upgrade tomcat to 8.5 but need to know if this issue is fixed in tomcat
8.5 or is it altogether new issue.

Could you please take look at this issue on priority and help us?

Regards,
Shailesh

-- 
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

Reply via email to