[Dxspider-support] Telnet connection keep alive

Pascal Pascal
Sun Nov 13 22:06:22 GMT 2016


The problem is that NAT and firewall don't know when a connection is ended.
Therefore they work with a timeout.
After the last packet is been send it starts a timeout.
When no data is received/send anymore in the timeout time,
the NAT/firewall function assumes the connection is ended, and forgets 
the session.
That is the reason why your connection is lost.
When you use Linux for NAT/firewall, you can make a rule for port 9000
(or whatever port you use for connecting the cluster), and give it a 
long timeout time.

On home routers/firewalls I am afraid there are no options to do this.
Greetings,

Pascal PB1SAM

On 11/13/2016 10:54 PM, Rene Olsen via Dxspider-support wrote:
> Hi.
>
> Some home routers have settings that are not very friendly to low traffic connections.
>
> Some of them have a timeout of just 5 minutes, and tears down the connection of there
> hasn't been a noticeble amount of trafic on the connection.
>
> Not even a CR and a prompt back from spider, is enough traffic to maintain the connection
> open.
>
> I have no idea why they do this. But since it is home routers, it is probably a question about
> money and ressources in the router.
>
> I had a linksys router once, and with the default firmware I wasn't able to change the timeout
> setting, and it was just not able to maintain the connection to the cluser, when there was very
> low number of spots.
>
> After an update to some third party firmware, I was able to change the timeout setting, and
> since then I didn't have any problems at all.
>
> So in my opinion most diconnect issues, can be explained by the router in use.
>
> And its not like the router tells both ends, that the connection has been closed. No, it just
> removes it from the NAT table and thats it.
>
> Of course there is good and bad routers out there. I can only talk from my own experiences
> with home routers in the past.
>
> Vy 73 de René / OZ1LQH




More information about the Dxspider-support mailing list