Thank you for your interest in this problem.
We restrict multiple access to the same application instance on our service.
We operate a chat using fms. We experience, from time to time, an application
instance failure which results in the entire room participants being
disconnected. The application instance however still shows the connections to
the room at time of this failure. The problem starts to compound itself as
people try to access the room ( connect to the application instance ). After
the start of this problem when we look in the admin console the live log for
the application instance does not show any new connection attempts nor accept
or failure to connect but the application instance registers the connection
against the total connect to the application instance. People are unable to
connect to this room. As they continue to try they increase the connection
number registered for this application instance until the point that the
application instance total connected takes us past our license limit. The
application instance seems to register and collect the connections even though
there is no connection to the application instance. We then start to see the
logging of the license limit exceeded.
Please note that there are no error/warning messages generated in logs that
indicate anything has failed. We can only determine that a failure once we
start to experience the symptoms described for the application instance.
We have experienced this issue in the past with fcs. It was not until the last
update of fcs that the probem went away. However, fms 2.0.0 onward has always
had this issue for us.
The problem with this issue is that it is difficult to create and study. We
can go 1 hour and crash or we may have success for over two weeks.
In an attempt to search for a solution to the problem we have started
focussing on this comment in Adobe ?TechNote: http://www.yqcomputer.com/
?In Server.xml, find the ResourceLimits node (around line 181). It contains
the Connector node (around line 203). Within the HTTP node (around line 205),
create a new line and insert MaxConnectionThreads and give it an appropriate
The default is 10.?
We have found that by adding the change to our HTTP node that we have lessened
the frequency of our problem. Also, we have taken all remoting away from the
time of connection. We now accept all connections and then do our work after
Another step we have taken to minimize the impact of this problem is to define
the scope as ?inst? in the application.xml. By creating a separate fmscore.exe
for each application instance it gives us the ability to correct a problem with
a faulting application instances without disturbing our entire server traffic.
Any further insight into this problem would be greatly appreciated.