I'm seeing a problem with random error disabled ports occurring during workstation start ups.
The problem is not confined to any one brand or model of NIC card, any one switch or switch port module, and the cabling system with patch cord attached has been tested and certified good.
The ports and workstations are all configured for autonegotiate and the problem never happens in the same place twice.
How can I troubleshoot this ?
Can the NICs be sending spurious electrical signals to the port when the workstation first powers on and that causes the disabled port condition ?
Any insights into this issue would be very greatly appreciated.
Are you seeing any late collision errors, or negotiations to half-duplex on the switch ports? Use sh int to check this. If you are seeing these errors, they are easy to correct by hard-coding speed and duplex settings.
Are you having any duplex mismatches?
Late collisions are an issue themselves, indicative of a problem. From Cisco TAC:
"Late collisions are usually physical layer issues. Its an indication of faulty hardware (NIC, cable, or switch port) or a duplex mismatch.
A late collision is a collision which occurs 'late' in the packet - in other words, sometime towards the end of the transmission of the packet.
A normal collision tends to happen in the first part of the packet (i.e. before 64bytes), and therefore is a simple coincidence - two workstations on the shared media trying to transmit at the same time.
A late collision means that everyone should have been able to see that some device was transmitting, but inspite of that some other device transmits anyway. In the case of
a duplex mismatch, the side which is half duplex sends the frame but the
full duplex side doesn't think it needs to wait and sends its frame as
well. The half duplex side then reports a late collision.
It may also be at times that if the interfaces is hard coded full on one side and the other side is auto the resultant will not be full.
Please do try to hard code the duplex as full and speed to 1000 on both the interfaces."
PS: TAC also reported to me in a separate exchange that it might be best to set the SPEED setting before the DUPLEX setting.
In our case, once we hard coded SPEED and DUPLEX on both sides, all errors stopped and throughput increased significantly.