Network Engineering Stack Exchange is a question and answer site for network engineers. It only takes a minute to sign up.
Sign up to join this community
I noticed gradually increasing input errors for a FastEthernet interface which is connected to an end-user Windows PC. The default setting for all interfaces is no-negotiate. The Foundry switch will use 100Mbps/Full if no-negotiate is configured.
The odd thing is that if I enable autonegotiation, the increasing input errors will stop occurring.
Is there any reason for this, and would there be a way to resolve the issue if I wish to configure the interface to use no-negotiate.
If you disable Auto Negotiation (AN) you need to make sure that both sides are configured in exactly the same way. There's isn't any point in doing that manually, actually, so you should have AN active at all times.
If the duplex settings of the link partners don't match you've created a duplex mismatch. The half-duplex side detects massive collisions and late collisions, the full-duplex side detects runts and FCS errors. The link works only rudimentarily and performs very poorly.
The classic reason for a duplex mismatch is forcing one side to full duplex while keeping the other side set to Auto Negotiation. Failing AN, that side falls back to half duplex, creating the mismatch.
Therefore, do not disable Auto Negotiation generally. AN is mandatory for Gigabit Ethernet, so disabling it causes a GE link to run 100 Mbit/s at best.
By manually setting the speed and duplex on one side with automatic detection on the other side, the side with automatic speed will detect (not negotiate) the correct speed.
For the duplex, negotiation will fail. This results in the side with automatic duplex set to the default duplex for the speed. The default duplex for 1 Gb is FULL, and for 10 and 100 Mb is HALF.
Your scenario results in a duplex mismatch which causes several problems that will slow the link to a crawl.
Cisco has a nice description in this document
