In my test environment I have installed a basic setup:
1 TIE server
2 DXL Brokers, configured as a hub named "Hub1"
1 Windows 10 client
1 Windows DC for Windows accounts and DNS
What I do not understand/get working is failover of the Windows 10 and TIE server DXL connection when doing an hard shutdown on the DXL broker they are connected to. When doing a kind "shutdown -h now" on the DXL broker the DXL clients on the Windows 10 and TIE server do a seamless failover to the other DXL broker. But when unplugging the power from the DXL broker the clients are having a hard time to connect to the other broker (I can not find a way to force a reconnect either). They succeed after some time (maybe 30 minutes of so it feels like). If I restart the TIE server it gets a connection to the started DXL broker. Until the TIE server has a working DXL connection it is not possible to see the TIE reputation in ePO and I guess the clients can not get it either.
Is this an expected behavior or is it possible to configure or build a DXL architecture to avoid this?
The default keepalive interval for DXL Clients' connection to a DXL Broker is ~30 minutes. This means that if a Broker goes offline without being able to send a shutdown notification to the fabric, the connected DXL Clients may take up to 30 minutes to become aware that the Broker is unavailable.
To directly answer your question: This 30 minute interval is within the expected behavior of a Client.
It is possible to modify the keepalive interval, but it is not recommended unless it is known for certain that there will be very few DXL Clients on the DXL fabric. Shortening the interval can greatly increase the amount of network traffic caused by the DXL Clients due to scaling issues (normal fabrics tend to have several thousand Clients).
I hope that helps clear things up!