0

We are investigating an issue on Windows Server 2019. It runs a Windows Service that accepts TCP/IP connections (TCP/IP service bound to IP address 0.0.0.0, so listening on all network adapters). External devices connect to the server via a public domain name / IP address. This has been working fine for a long time.

A few days ago, device connections have started dropping out on occasions and thereafter devices seem to be unable to reconnect, while other connections remain alive and unaffected. While in this faulty state, connections to the Windows Service on the actual machine (using a test application) are still accepted when connecting to localhost, but connection attempts via public IP address or domain name fail. Strangely, the public IP address and domain name can be pinged successfully on that machine; the Windows service just won't accept the connection.

The Windows service is based on the .NET 6 framework, and once it starts listening on its assigned port, it does not stop. Since a connection to localhost still works, that would mean that, in general, the application is still alive in the listen/accept loop. It does not change the listen port or the bound IP address at runtime.

Restarting the service does restore connectivity for a while until some devices drop out again.

There have not been any Windows updates recently; the last update was well before problems started occurring. I'm not aware of any changes in the network configuration (or any other changes for that matter, at least for now).

Has anyone come across such a problem?

What to look at next?

4
  • What to look at next? What are the number and duration of connections?
    – Greg Askew
    Oct 26 at 10:58
  • There are less than 50 connections and they are long-lived, so normally all connections are up without continuous disconnect/connect.
    – J.R.
    Oct 26 at 11:24
  • I would run a netmon packet capture to confirm the packets are actually making it to the host. And to confirm it isn't responding, or what the response is.
    – Greg Askew
    Oct 26 at 11:28
  • Thank you for your suggestion. The problem was solved by re-building the virtual machine. Unfortunately the root cause will never be known.
    – J.R.
    Oct 26 at 20:04

0

You must log in to answer this question.

Browse other questions tagged .