[PaulNSW]

I understand this may not be a Maxwell issue, but I was hoping someone could help!

We have a set of rendernodes on a 192.168.101.0/24 subnet and during tight deadlines, we add some extra nodes from another subnet, 192.168.102.0/24 joining the manager on the 101.0

They join up fine, report to the manager, render happily - everything is fine. However, if the 102.0 node is left idle for 15 minutes, the connection to the manager is dropped. If I restart the node software, it will join up.

I've looked in the power settings on the computer (Windows 7 x64), it's set to always on. I've unchecked powersaving on the network adapter and I've disabled the Windows firewall on all machines.

If I put the same machine on the 101.0 subnet instead, it remains connected fine. Which makes me think it's something to do with our internal routing. The only tip I saw here was about fragmented datapackets, and that's enabled.

Does anyone have any other ideas?

Thanks for any help!

[PaulNSW]

Ok, I just had a look in our firewall/router

It seems there is a default TCP time-out of 15 minutes. I can increase it, but the firewall complains:

If a connection to a remote server remains idle for more than five minutes, the SonicWALL closes the connection. Without this timeout, Internet connections could stay open indefinitely, creating potential security holes. Setting excessively long connection time-outs will slow the reclamation of stale resources, and in extreme cases could lead to exhaustion of the connection cache

Maxwell makers, would it be possible to enable a bit more "chatter" between the manager and render node, rather than just opening a connection? Maybe every 10 minutes or so an "are you still there?" packet could be sent.