A software development team reported random connection drops between their application server and the database. The database administrators confirmed the database itself ran correctly, and the server’s resource usage was low. Reviewing the network interface statistics on the application server, I found a high count of dropped packets and cyclic redundancy check (CRC) errors.

My charge was plain: find the cause of the dropped packets and restore the connection between the servers.

I began a packet capture on both servers to watch the traffic, and examined the physical network path. The servers connected to different access switches. I checked the port statistics on both switches; the port serving the application server showed many CRC errors as well, which pointed to a fault at the physical layer. Inspecting the patch cable between server and switch, I found a sharp bend near the connector. I replaced the damaged cable with a new category 6A patch cable.

I cleared the port statistics on switch and server alike and watched the connection for 30 minutes. The CRC errors ceased to climb, and the packet captures showed clean data transfers. The development team confirmed that the connection drops had stopped entirely.