Networks

Ask something
Discussion Locked

Lots of input/crc errors

0 helpful
7 replies

by daveg

Closed to new repliesThis discussion is closed. It stays readable, but it takes no new replies.
We have a 3C982-TXM 10/100 PCI dual port NIC with dynamic access technology on a Windows NT cluster server. The NIC is configured for 100Mb/Full duplex as well as the Cisco Catalyst 3500 series switch that it is connected to. Usually this configuration solves any crc errors picked up by the switch. We've gathered 1200+ crc errors on one port and 550+ crc errors on the other port and these errors correspond to the input errors. Where could these errors be coming from and should I be concernedabout them?

7 replies

by maxwell edison

A CRC error could be indicative of some hardware problem. These hardware problems can include a bad hard drive controller, a bad hard drive, or a bad cable.

For optimal data transfers, you should make sure you a have an IDE cable less than 17 inches in length. Cables longer than 18 inches can cause CRC and other errors due to the speed of data being transferred. Also, make certain that you do not have the cable tied down or wrapped back and forth. While this makes a box easier to work with, this practice could cause CRC and related problems due to the effects of electro-magnetic feedback in the cable. Another thing to do is keep noisy electronic equipment away from the drive and cables. Also, make sure that the IDE drive has an inch of space between it and other hard drives and fans.

Maxwell

by shanghai sam

This cluster server has an adaptec 294x SCSI connection to an external RAID controller. Previous to this NIC, we had a DEC NIC which registered no crc's. We went to the 3com card for the dynamic access technology but there seems to be some communication problems between the NIC and the switch, maybe a compatibility issue between cisco and 3com. I am not sure.

by etherealmind

The most likely cause of this error is a speed / duplex mismatch between the 3Com adapter and the Cisco Catalyst switch.

The recommendation would be to configure the Catalyst switch to 100Base at Full Duplex and then cold start your server.

The Fast Ethernet autonegotiation is not inherently reliable and often the two connections will set themselves to wrong settings. For example, the server would select 100BaseT Half duplex, and the switch would select 100BaseT Full. This will lead to very high numbers of CRC errors and other input errors on both ends of the connection and very poor overall throughput numbers.

For more information go to Charles Spurgeon's Ethernet site for more detail than you ever wanted to know. DO a search at Google on the word "Spurgeon".

It also just remotely possible that if this does not fix the problem that you have a bad patch lead. You should change it just to be sure.

Rgds

Greg

by daveg

Patch lead already checked, and the speed/duplex for both ends are set manually to 100/full. This is why at this time I am leaning toward a compatibility issue between 3com and Cisco. I will check out the Spurgeon link though, thanks.

by etherealmind

If you are using WinNT 4.0 or Win2K you should try 100Base Half not 100BaseFull. I have experienced much weirdness fron WinNT at 100BaseFull and do not usually recommend it.

Note also that the relative numbers of errors. If you have 12 billion packets and only 1200 CRC errors, this is well within expected norms. Rule of thumb is about 0.1% would be OK by me.

Greg

by daveg

I guess that 28 billion input packets and 2806 CRC errors is in the expected range. Thanks

This discussion is closed to new replies.