Where to look for errors

[Scott Frye_1]

HP-UX 11.0. I have been told our network people are seeing errors from a switch. My NIC is set up as such:
Autoneg: OFF
Speed: 100
Duplex Mode: Full.
The Cisco switch is set up the same. I was asked if I could see any "overruns or CRC errors". I have no idea where to look for such a thing. Can someone point me in the right direction. They are thinking I might have a faulty NIC or need an upgraded driver.

[Pete Randall · Accepted solution]

Scott,

Have a look at lanadmin output. It gives you stats on discards, errors, collisions and the like.


Pete

[Scott Frye_1]

Specifically, I am interested in understanding what these errors mean.

Port Align-Err FCS-Err Xmit-Err Rcv-Err UnderSize
----- ---------- ---------- ---------- ---------- ---------
3/7 75147 67217 0 0 0

[rick jones]

those errors suggest something non-trivialy amis at the physical/datalink level. Triple check that the switch port is indeed set to FD - ass-u-me-ing the stats you give are from lanadmin -g mibstats on your system, FCS errors can be an indication of a duplex mismatch. It could perhaps be a bad cable, bad NIC, or bad switch port.

Ass-u-me-ing again that all the ports on the switch are set to 100FD, you could try moving the cable to another port on the switch. (DONT DO THAT IF YOU DONT KNOW ALL PORTS ARE 100FD)

[Scott Frye_1]

These errors are from the switch side. When I do the lanadmin -d my errors are 0. We have double and tripple checked to make sure the switch is set at full duplex. So in trouble shooting this, would you 1. switch out he nic, 2. update a driver (if available), 3. change ports on the switch? Please help me in determining my next steps.

Thank you very much.

Scott

[Brian Hackley]

Scott,

Don't forget to first isolate the cable. Could be something in one of the cable segments, patch panel, etc. One way to do this is to swap the cable ends of 2 network cards on the HP and the switch end...if the problem follows the cable...

HTH, Brian Hackley

[rick jones]

I would start by swapping the ports with one that was "known good." If there are still errors on the "known good" port then you should try swapping the cable. It doesn't really matter the order in which you do them, just that sometimes it could be rather difficult to run a new cable, so I try to put that one off - just a matter of personal preference.

That there are no errors on the NIC side, and errors on the switch side makes me (for no specific reason really) wonder about the switch port.

I do not think that changing a driver would affect FCS errors - those are at the hardware level. I doubt it would _hurt_ to update the driver (or the switch firmware for that matter) but I wouldn't expect it to change much. Still, if the revs are behind, this might provide a decent pretext for the downtime.

[Scott Frye_1]

Very good information from all. We are scheduling a downtime to upgrade the memory in the machine. At that time we are planning on switching the cable out and puttnig it into a new port on the switch (Two birds with one stone). If that does not resolve the issue, would you all concur that it could be an issue with my NIC card?

[rick jones]

The only issue with two birds with one stone is you won't know which bird was hitting you with its droppings (as it were). So, if things start to work, you won't know if it was the port, or the cable that was bad, so you don't know if you need to toss the cable (after cutting the ends off to make sure no one else uses it) or never use that port on the switch.

Indeed though if after you've swapped cables and used another port (ass-u-me-ing you've used a _known_ good cable and a _known_ good port) the next suspect would be the NIC. If you were on 11.11 (11i) and on the right hardware you could try that experiment without having to bring-down the box :) (At least I _think_ the 100BT NICs have OnLine Replacement (OLR) support on 11i v1 and later)