A network connection that should be running at 1 Gbps but consistently negotiates at 100 Mbps is a classic physical-layer troubleshooting problem. The switches may both support Gigabit Ethernet, the configuration may look correct, and yet the link refuses to operate at the expected speed. That's exactly what happened during this troubleshooting job. Rather than immediately replacing the switches or changing configuration settings, I started at the physical layer and worked through the cable path. The investigation eventually uncovered a hidden Ethernet connection that was the likely cause of the problem.
The first clue was the negotiated link speed. Both switches had Gigabit-capable Ethernet ports, but the connection between them was showing 100 Mbps full duplex. When a Gigabit Ethernet connection falls back to 100 Mbps, the cabling should be one of the first things checked. 100BASE-TX uses two twisted pairs, while 1000BASE-T requires all four pairs to be working correctly. A damaged conductor, bad termination, excessive untwisting, poor connector, damaged cable, or other physical-layer problem can therefore prevent a Gigabit link from negotiating correctly. Checking the switch interface statistics is also useful because physical problems can produce errors, dropped packets, or other indications that something isn't right.
I connected a cable tester to the connection to determine whether the problem was actually in the cable. The tester confirmed that the connection wasn't performing as expected at Gigabit speeds. I then used a TDR (Time Domain Reflectometer) to investigate the cable run in more detail. TDR testing works by sending a signal down the cable and measuring reflections caused by changes in impedance. Depending on the capabilities of the tester, this can provide an estimated distance to a fault or other significant event in the cable. In this case, the measurement provided an important clue because the problem appeared to be located somewhere around the middle of the cable run.
That sent me looking along the actual cable path, and eventually I found something that wasn't shown on the network documentation: an inline RJ45 Ethernet coupler hidden in the run. The connection had apparently been working for a long time, so it would have been easy to overlook. An Ethernet coupler isn't automatically a problem, but every additional connection introduces another potential failure point. A damaged connector, poor termination, corrosion, mechanical stress, or a component that isn't suitable for the required Ethernet speed can affect the electrical performance of the link. Since Gigabit Ethernet uses all four twisted pairs, a connection that is marginal may still allow 100 Mbps operation while preventing a reliable 1 Gbps link.
To confirm the diagnosis, I removed the questionable connection and temporarily used a small Gigabit Ethernet switch as a test point. The result was immediate: the connection negotiated at 1 Gbps instead of 100 Mbps. I then cleared the interface counters and allowed the connection to operate for a period of time while monitoring it. The link remained at Gigabit speed and the interfaces did not begin accumulating errors. This was a useful confirmation that the switches themselves were not the problem and that the fault was somewhere in the original physical connection. A temporary test configuration like this can be very useful during troubleshooting because it allows you to change one variable at a time and see whether the behavior changes.
The final solution should not necessarily be to leave a temporary switch sitting in the middle of the cable run. Once the faulty section has been identified, the better approach is to eliminate the questionable connection or replace it with an appropriate, properly rated component. In many installations, a continuous cable run is preferable because it removes unnecessary connection points and makes future troubleshooting easier. If a connector or coupler is required, it should be suitable for the Ethernet speed being deployed and installed correctly. After making the repair, test the cable again and verify the negotiated link speed and interface statistics on the network equipment.
This troubleshooting job is a good reminder that a 100 Mbps link on Gigabit-capable equipment doesn't necessarily mean there's a switch configuration problem. The physical layer can often provide the answer. In this case, the investigation went from the switch interface, to cable testing, to TDR measurements, and finally to a hidden RJ45 coupler. Each test narrowed down the possibilities until the problem became obvious. That's the approach I prefer with network troubleshooting: don't guess, test each part of the path, and use the results to decide what to check next. A few minutes spent testing the cable can save a lot of time chasing problems that don't actually exist in the switches or configuration.
