Time out on Established Connections

[Sundeep Kumar]

We have Cloverleaf listeners ports (threads) LAB_IN and EMR_IN and remote client connects to these ports over VPN and send transactions. Sometimes

the VPN just drops but at our end we show established connection on those ports (netstat -a)

meanwhile VPN comes back the client tries to connect but we are not listening anymore ie

connection show established! not listen so remote clients are not able to connect back till

we stop and start the thread, IS there a way out ? Can we timeout a connection on a listener thread after a preiod of inactivity

and go back to LISTEN state on that port

Topic

[Bill Bertera]

you can use multi-server ports. that way even though the first connection still thinks it up, it will allow a second connection. Your OS should eventually kill the first connection.

[John Hamilton]

There are numerous posts that deal with this VPN issue on this site.

I suggest you look at some of them and chose the preferred method.

My preferred method has been to get the network guy to fix the VPN.

Failing that the vendor has been able to either dis-connect after write or send a keep alive message to keep the VPN up and running.

Running a Multi port if you are using and ACK style message requires special coding so I would avoid that unless you understand what it require and how to make it work.

[Amol Kulkarni]

Sundeep, try setting the keepalive setting to 1800 = 15 minutes if you are on an AIX server.

To determine current keepalive setting:  no -a | grep tcp_keepidle.

[Gary Atkinson]

I just went through a similar issue, but cloverleaf was the sender.  I just kept complaining to the network guys and they worked it out.  I did set-up recovery procs on the thread to re-send if an ack is not received.  VPN’s are pain  👿

[David Bolin]

Gary Atkinson wrote:

I just went through a similar issue, but cloverleaf was the sender.

[Gary Atkinson]

The vendor set-up some kind of “ping script”; unfortuantely I don’t know the specifics.  Our network guy and the vender network guy matched up “the settings of the vpn”.  I believe what fixed the issue was the ping script.  I don’t know if it was specifically just “ping”.  I just kept pushing the issue that is was not cloverleaf.  What I did was turn logging up and track everthing that was going on.  I noticed that every 2 hrs and 13 minutes the process log posted a connection-time out.  Why it was always in that interval I don’t know.  I set-up the re-send proc to keep sending the message until an ack was received.  I set my time out to 30 seconds.  Since logging was on I could show the vendor when the engine received the message and time the message was re-sent.

[Author]

Replies