Poor network xfer speed, FCS errors

[Ed Loehr]

I see FCS errors in lanadmin and abysmal network transmission speed to/from this box on our lan. Duplex and speeds match the other ends (100, full). Noticed on another (linux) box having same trouble that it has flow-control enabled on Broadcom BCM5701 card, possibly a know troublespot. Wondering if this card has flow-control on, and how to troubleshoot this further?


root@max:/ $uname -a
HP-UX max B.11.23 U ia64 1562128462 unlimited-user license
root@max:/ $lanscan
Hardware Station Crd Hdw Net-Interface NM MAC HP-DLPI DLPI
Path Address In# State NamePPA ID Type Support Mjr#
0/0/3/0 0x00306E4C1686 0 UP lan0 snap0 1 ETHER Yes 119
0/1/2/0 0x00306E4C1687 1 UP lan1 snap1 2 ETHER Yes 119
root@max:/ $lanadmin -g mibstats 0

LAN INTERFACE STATUS DISPLAY
Fri, Sep 3,2004 14:59:22

PPA Number = 0
Description = lan0 Intel PCI Pro 10/100Tx Server Adapter [100BASE-TX,FD,MANUA
Type (value) = ethernet-csmacd(6)
MTU Size = 1500
Speed = 100000000
Station Address = 0x00306e4c1686
Administration Status (value) = up(1)
Operation Status (value) = up(1)
Last Change = 230930318
Inbound Octets = 3934626063
Inbound Unicast Packets = 28759947
Inbound Non-Unicast Packets = 5415888
Inbound Discards = 152059
Inbound Errors = 25816
Inbound Unknown Protocols = 820886
Outbound Octets = 2547486556
Outbound Unicast Packets = 23049190
Outbound Non-Unicast Packets = 47876
Outbound Discards = 0
Outbound Errors = 0
Outbound Queue Length = 0
Specific = 655367

Ethernet-like Statistics Group

Index = 1
Alignment Errors = 28
FCS Errors = 38158
Single Collision Frames = 0
Multiple Collision Frames = 0
Deferred Transmissions = 0
Late Collisions = 0
Excessive Collisions = 0
Internal MAC Transmit Errors = 0
Carrier Sense Errors = 0
Frames Too Long = 0
Internal MAC Receive Errors = 0

[rick jones · Accepted solution]

If I had to guess, I would say that you have hardcoded the NIC to Full-Duplex (note the interface description field) but the switch port is set to Autonegotiate.

When one side is hardcoded, and the other is at Auto, the Auto will "fail." When auto "fails" the side(s) doing auto are required by the spec to go into half-duplex mode - the presumption being one is communicting with old legacy half-duplex only equipment.

Thus you get a duplex mismatch - one sign of which are FCS errors when the NIC is in full-duplex (or _late_ collisions when the interface is in half-duplex.)

I would suggest you disable the hardcode and see if that fixes things. If that does not, _then_ consider hardcoding both the NIC and the switch to the _same_ setting. And then to prevent future issues, hardcode _all_ the switch ports to that value, along with _all_ the other systems. Otherwise, the first time someone goes to "neaten" the cables at the switch...

...my preference is to stick with auto everywhere until proven (not presumed, but _proven_) broken.

[Ed Loehr]

Pretty good guess. Turns out this was apparently a bad NIC that was stuck in essentially a hard-coded state (maybe full duplex). I won't risk naming the manufacturer as my memory since has grown fuzzy.