Connection of clients and number of licenses.
All days I touch the limit of licenses, because is impossible for me to remove old connections; even if the connections are still working, is impossible for me (sometimes) to remove from the list of connected clients! Is impossible to implement an automatic job (e.g. in the night) that automatically disconnect all clients?
I found this problem too. A customer of mine has some networking problems so sometimes I find in the log an error saying ‘could not connect to client xx-xx-xx-xx-xx-xx: no route to host’, this error leads to very bad things:
1- sometimes the client show the ‘Connection to the server lost’ dialog; (this is fine)
2- Sometimes I see from the admin pages that last ‘alive check’ for the client is really old and sometimes the client is already closed but I cannot kill it and it keeps one of my licences busy; (this is bad)
3- sometimes the client just doesn’t notice but data broadcast stop working for that client until client restart; (this is really bad!!)
I know it depends on network problems that should’t happen but anyway point 3 is really scarying.
Could Servoy react to network problems in a better way?
Maybe I should post this in Issues and Bugs?
Servoy version R2 2.2-build 328, repository version 25
os.name=Linux
os.version=2.6.8-24-smp
os.arch=i386
java.vm.name=Java HotSpot™ Client VM
java.vm.version=1.4.2_07-b05
java.vm.info=mixed mode
java.vm.vendor=Sun Microsystems Inc.
Please run Servoy Server with -DSTACKTRACE=true and send me the log when you try kill a client from the Servoy Admin pages and it does not disappear within 10 minutes (note: a client need not disappear immediately, because a connection attempt is made to the client to shutdown gracefully).
I have not seen any of these problems before, and they should not happen. When you kill a client it should disappear 1) immediately if it still reachable, 2) within 10 minutes if it is unreachable. Furthermore broadcasts that do not reach the client because of network errors are retried, so the client should get them (with a delay). If the client is no longer registered with the server because the network outage is too long, the client should notice and reregister itself.
So in all cases where these errors are appearing, please send me the server log (with -DSTACKTRACE=true), and if possible the timestamp of when approximately the failure occurred.
Ok, I’ll send you a stacktrace. I’m very concerned about this one.