Are your Domino sessions between servers or from client to server going away? Over the last months I’ve seen this many times and had a hard time tracing it down.

In a private forum for Business Partners, we’ve had some success at drawing the attention of the right people from IBM to the issue but these things never resolve quickly. Even if the problem was instantly diagnosed, the cycle for fixes tends to be about four months as they must be fit into the build/test cycle. It’s an approach that avoids problems by reducing the repeated patch fixes if you try to tackle each – but in the meantime it can be painful. Here’s what I’ve been able to put together so far in my own lab:

I see this problem occurring most often when Domino 6.5x is running on a Win/32 platform with a single network interface card that has more than one network address tied to it (multi-homed) and the connection your making is not to the first address listed on the card. Small shops with a single server that handles connections both inside and outside the firewall seem to be most vulnerable.

What happens is a connection get established when you first connect, but then after some amount of time not using the connection, you try to save a document or open something else on the server and the connection times out with a ‘server no longer responding’ kind of error. If you hit ctrl-break during the connection attempt to invalidate the current connection then try again it will connect cleanly.

What remains unanswered for me so far is:

1. Why is the connection timeout not invalidating the connection when it fails, so that the next attempt works even without a ctrl-break? This would clearly help the issue for unattended connections or servers.
2. Why is the connection timing out in the first place? Is this a function of something on the network? I suspect not. I suspect this has to do with one of the Windows patches that attempt to avoid DOS vulnerabilities like a SYN-flood by closing sessions which are unused.
3. Why can’t the client side determine more quickly that the connection is invalid and create a new one?

Anyone else seeing this?