Blade server cannot communicate through CPM ports

If a blade server cannot communicate to the network through a copper pass-thru module (CPM), verify that the pass-thru module is functioning, that the expansion cards in the blade server are compatible, and that the network switch setting match the settings on the copper pass-thru module.

Problem

A blade server is unable to communicate through a CPM.

Investigation

The Copper Pass-thru module (CPM) is a gigabit-only device; connections to 10/100 switches will not work. All blade server-to-blade server connectivity requires the CPM to be connected to a properly configured external switch. The external switch ports must be copper gigabit ports configured for autonegotiation. All other configurations are unsupported and can result in errors that make it appear that the CPM is not functioning correctly.

The CPM does not have switch functionality, which requires that you be able to see the configuration of the external switch when troubleshooting CPM connectivity problems. In addition, Serial over LAN (SoL) will not work if there is a CPM in I/O-bay 1.

The following link indications are available for isolating a connection problem:
  • The external (upstream) switch likely provides a link indication. It indicates a link in the conventional networking sense of layer 1 link being up between the CPM and the upstream switch. Even with all blade servers removed from the BladeCenter S chassis, the upstream (external) switch link light will be on if the external cables are properly connected.
  • The advanced management module provides internal and external link status (from the CPM perspective). If the external link indicates up, it means that both sides of the connection (external switch to CPM and CPM to blade server) have successfully negotiated the speed and duplex mode and that carrier is detected from the blade for that connection. If no internal signal is detected, the external indication will not provide an indication (it will not indicate a connection whether one exists or not).

    The internal link indication does not indicate a successful negotiation of speed and duplex. It only tells you that a blade server is in the slot, and some electrical current has been detected. For example, if you boot a blade server into BIOS and disable the NIC, the link still shows up, even though speed and duplex have obviously not be successfully negotiated. (The one exception to this is that if the system is booting Linux, and the NIC driver module is removed, the link will show down after about 15 minutes).

    Unlike the advanced management module external link status, the internal link status, as seen from the advanced management module Web interface, is independent of the external link and is accurate with and without an external connection. Link up indicates that the Ethernet port, Ethernet driver, cable, and upstream port and blades are configured correctly for layer 1 of the OSI model.

  • The operating system on the blade server provides link indication from the Ethernet port.
  • The LEDs on the CPM are equivalent to the advanced management module viewable external link status. Only one physical CPM LED is associated with each blade connection. When that CPM link LED lights, both the internal (CPM to blade) and external (CPM to switch) connections recognize electrical signals are connected.

No connectivity on any ports

If none of the blade servers in the BladeCenter S chassis can communicate through the CPM:
  1. Check POST codes for the CPM. Resolve any POST errors that are found.
  2. Make sure that external ports are enabled for the CPM through the advanced management module Web interface.
  3. If there is another Ethernet module that is working, swap this CPM with the other Ethernet module to verify whether the problem follows the module or stays with the I/O module bay. If it stays with the I/O module bay, see the Ethernet troubleshooting document.

Connectivity on some ports but not others

In the BladeCenter S chassis, all blade server Ethernet ports are connected to I/O module bay 1. The mapping of blade server ports to the CPM ports are as follows:
  • Switch Port 1 - Cable 1A - blade server ethernet port 1A
  • Switch Port 2 - Cable 1B - blade server ethernet port 2A
  • Switch Port 3 - Cable 1C - blade server ethernet port 3A
  • Switch Port 4 - Cable 1D - blade server ethernet port 4A
  • Switch Port 5 - Cable 1E - blade server ethernet port 5A
  • Switch Port 6 - Cable 2A - blade server ethernet port 6A
  • Switch Port 7 - Cable 2B - reserved
  • Switch Port 8 - Cable 2C - blade server ethernet port 1B
  • Switch Port 9 - Cable 2D - blade server ethernet port 6B
  • Switch Port 10 - Cable 2E - reserved
  • Switch Port 11 - Cable 3A - blade server ethernet port 4B
  • Switch Port 12 - Cable 3B - blade server ethernet port 5B
  • Switch Port 13 - Cable 3C - blade server ethernet port 2B
  • Switch Port 14 - Cable 3D - blade server ethernet port 3B
  • Switch Port 15 - Cable 3E - reserved
If some ports on the CPM work but others do not:
  1. Make sure the CPM cables are oriented correctly. These cables are not keyed and can be connected upside down. An upside down connection results in internal link down on some ports and link up on others. The cables join the metal cable connector at an angle. When the cable is properly oriented, the cable will go up from the metal connector.
  2. Verify that all external cables have links indicated by the upstream (external) switch. If some do not, examine the switch configuration to ensure that the upstream switch ports support gigabit and are configured for autonegotiation.
  3. Examine the statistics for the upstream switch and verify the speed and mode the link is running at now. The speed/duplex results should be 1000/full. If not, swap the RJ-45 connector between a working and non-working port to isolate this to a cable or upstream port problem.
  4. If the step above indicates that the octopus cable is at fault, move the cable between the three connectors on the CPM to determine whether the issue is with the cable or the connector. Replace the component that appears to be bad.
  5. Verify that the blade server Ethernet ports are correctly configured within the operating system and show link up. Remember that seeing internal link up does not, mean that the Ethernet port has a proper layer 1 link.
  6. Replace the CPM.
Parent topic: Troubleshooting I/O module problems