[APLO-140] CPU Spin state using SSL connection, connections not properly cleaned up.

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 1.0
• Component/s: apollo-broker
• Environment: apollo-1.0-0.20120124.204817.255; java version "1.6.0_30"; Java(TM) SE Runtime Environment (build 1.6.0_30-b12); Java HotSpot(TM) 64-Bit Server VM (build 20.5-b03, mixed mode)
• Assignee: Hiram R. Chirino
• Reporter: Massimo Paladin
• Created: 27/Jan/12 14:34
• Updated: 30/Jan/12 16:34
• Resolved: 30/Jan/12 16:34

Description

I can reproduce the problem doing the following:

• running stomp benchmark with the attached scenario
• run check_stomp in a loop against the broker on stomp+ssl (check_stomp in this case connect via ssl, drain the destination, send one message and check he receive that message back)
  the check_stomp command line /path..to/check_stomp --config /path/to/file.conf --broker stomp+ssl://host --destination /topic/test.stompssl
  to see what check_stomp does: http://svnweb.cern.ch/guest/MIG/trunk/mig-nagios-plugins/src/check_stomp

This has as result:

• very high cpu load remaining after stopping the tests, it remains high until you restart the broker
• apollo web console reporting some stomp+ssl connections open which are not appearing in netstat at all

I attach you a thread dump where I am sure that no clients could have been connected.

nestat output for stomp+ssl port:
tcp 0 0 0.0.0.0:61133 0.0.0.0:* LISTEN 25699/apollo

The screenshot about the web console is attached also, it shows the connections remaining in Apollo.