Idle Session Disconnection

[johnd↑↓]

Hi Folks.

I've noticed that connections through NAT - specifically connections to a remote SQL Server database - appear to be closed by the proxy if they remain idle for about 4 minutes.

Is this intended, and is there any way to modify this behaviour, or extend the timeout?

We are running 5.2.3.

[Pascal↑↓]

Is this intended, and is there any way to modify this behaviour, or extend the timeout?

Hi,

It is intended. Any session which does not have traffic will be disconnected after the timeout period. There are two ways to work around this. The best way is if your client application has some way of sending keep-alives. The second way is to add a port security action from the LAN to the Internet on the port(s) that the application will connect to. You can set the action to "Allow packet" and then control the timeout that way. There are three options, "Use Default", "Never Timeout" and a user specified timeout. I'd recommend using the user specified timeout, with what you would consider to be the optimal time.

[johnd↑↓]

OK, thanks Pascal.

[johnd↑↓]

I've now tried this, but it doesn't appear to have any effect.

LAN connections to Internet
Protocol : TCP
Port : 1433
Allow Packet
Custom timeout after 7200 seconds

Have I missed anything?

[johnd↑↓]

Hi,

Any ideas on this?

[Pascal↑↓]

I'll have to run it past QA. Double check that the port is actually 1433 (Can check that when a session is active), that the session is going through WinGate, and that it's not the remote server closing the connection.

[johnd↑↓]

Definitely the right port. I also get the same behaviour with MSN Messenger, and it's always in the region of 250 seconds, which makes me think the common factor is the proxy.

Now, this may not be relevant, but I'll mention it just in case. I exchanged some correspondence with your colleague Gene Soudlenkov regarding a problem we had with the Wingate firewall. There's something about our internet connection that confused it, and I'm currently running a version of the executable with an interim fix, that isn't yet in the official release. So, is it possible that whatever confused the firewall is also affecting the timeout behaviour?

[Pascal↑↓]

That may be likely. Was this the problem with the strange route entries ? (USB Modem ?)

[johnd↑↓]

Yes, that's the one.

[johnd↑↓]

Do you have any thoughts or feedback on this yet?

[Pascal↑↓]

Not yet, sorry. We're getting things setup in the QA lab, which might take until tomorrow or so.

[johnd↑↓]

OK, thanks Pascal.

[johnd↑↓]

Are you still progressing this Pascal?