Skip to main content
Contributor
April 28, 2008
Question

Idle connections freeze or lock RDP and MySQL

  • April 28, 2008
  • 2 replies
  • 2642 views
We have a problem with a HA FortiGate-300 cluster. The cluster is in transparant mode. We have found that if we leave a command line mysql client open and idle, it will become locked out after a couple of minutes. It doesn' t disconnect, it just hangs on the next query as the tcp connection has gone away. We see something similar with Remote Desktop connections through the FortiGate. If left idle for a while, the connection will go away. RDP will recover, but the desktop will be frozen for about a minute. As the FG is in transparant mode, I assume session time-outs don' t come into play, or do they?

    2 replies

    abelio
    SuperUser
    SuperUser
    April 29, 2008
    As the FG is in transparant mode, I assume session time-outs don' t come into play, or do they?
    Timeouts play even in TP mode, but you' re talking about couple of minutes, a very short timeout. Defaults are higher (3600 sec), so it seems to be another issue. Check table sessions in dashboard and play with ' config system session-ttl' CLI command for your protocols, i.e, for rdp standard 3389/tcp port:
      config system session-ttl          config port              edit 3389                  set timeout 7200              next          end  end  
    Contributor
    April 29, 2008
    Abel, You did however hit the nail on the head. Either the default changed or the previous admin (I inherited this setup from a takeover) had some interesting ideas...
      config system session-ttl      set default 300          config port              edit 22                  set timeout 3600              next          end  end  
    abelio
    SuperUser
    SuperUser
    April 29, 2008
    interesting, maybe your previous admin tried to react to ' The system has entered conserve mode' message for a overloaded box; there' s an kc' s article (http://kc.forticare.com/default.asp?id=1076) where one of the emergency settings was set default timeout to 300 sec. Or the emergency became permanent or you' ve inherited the box meanwhile; maybe the cluster was the way to solve the overload, or maybe earlier admin ' had some interesting ideas... ' best regards,