Collisions showing on my NIC
 

Greetings,
When I run ifconfig on my Linux server, I see a lot of collisions on the interface eth0. However, my Cisco switch is not registering any collisions. How do I check to make sure that I am not having problems on my intergrated NIC card. I am new with to the Linux As 2.1 environment, so please be gentle (-:

I just need some guidance what to do to test out the NIC/interface on my Linux server.

Thank you in advance.

JB


eth0 Link encap:Ethernet HWaddr 00:0B:DB:E6:F4:FD
inet addr:192.168.11.115 Bcast:192.168.255.255 Mask:255.255.0.0
UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
RX packets:338473 errors:0 dropped:0 overruns:0 frame:0
TX packets:118437 errors:0 dropped:0 overruns:0 carrier:0
collisions:582 txqueuelen:100
RX bytes:33389366 (31.8 Mb) TX bytes:57095808 (54.4 Mb)
Interrupt:17 Memory:eff10000-eff20000

lo Link encap:Local Loopback
inet addr:127.0.0.1 Mask:255.0.0.0
UP LOOPBACK RUNNING MTU:16436 Metric:1
RX packets:12145 errors:0 dropped:0 overruns:0 frame:0
TX packets:12145 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:0
RX bytes:11060701 (10.5 Mb) TX bytes:11060701 (10.5 Mb)

Just curious, but do you have the switch set to auto negotiate?

Clark,
No, I have the port set at 100Mbps Full-Duplex - I do know that Cisco switch doesn't like "auto negotiate" setting at all. But I did thought about setting it to "auto negotiate" though.

thanks for your response.

JB

Well, I see you were thinking the same thing I was in regards to the duplex setting. I myself am more inclined to trust what your switch is telling you. It's a 100 mbps full duplex link, so obviously a collision isn't possible in the typical sense of the word. I don't suppose you have a extra PCI NIC that you could throw in for giggles? The problem with the NIC is we just don't get much out of it in terms of information. Obvious question, so I think you have most likely tried it, but have you tried moving the server over to a new switch port? Are you actually seeing performance issues or are you just curious about the collisions? I couldn't blame if it were just curiosity, as I would be curious as well. Hmmm. Another thought is your patch cord. Have you tried a new patch cable?

No, I haven't tried to install another NIC or switch to another port, because this is a "production" server and I have to schedule a down time to do so. And yes, I have pinged another linux server continuously and I managed to kicked off a few users and including my self. The server stayed up and idled at 99.4% CPU and not mentioning that the server was very responsive at the console.

Currently, I have an intergrated NIC on my Dell server which is a bcm5700 model according to to lsmode command. I also found out that dell has a new updated driver 9/04 for this particular NIC. I might update this driver in the next couple of days or so.

This problem just recently appears for about a week now. I would like to find the bottom of this issues soon. I might switch out my network cable tomorrow and see if that would help.

I do have another PCI NIC that I can throw it in there just for kick...What's the procedures to do that since this is not a Windows OS, so not sure if linux has the plug& pray capabilities.

How should I attempt this procedure? Thanks!!

JB

The production server aspect does always throw some nice challenges in there doesn't it? However, in order to really troubleshoot the issue, I anticipate you'll have to schedule a time in those off hours to really get down to brass tacks with this thing. Again, I trust your output from your switch more than I do the NIC. If it were me I would:

1. Schedule some down time.
2. Move the server over to a new switch port and check for collisions again.
3. If you still have collisions, replace patch cable.
4. Check for collisions again.
5. If you still have collisions, get out your spare NIC and try it again.

In regards to the detection of the NIC, we really want to examine a couple of fairly simple issues. What kind of NIC? Make sure it is supported first of course. Odds are it is. So get your make and model and do a little digging there. Then if it is, depending on how your kernel is configured, it might get detected right away. It depends on if you have the module for that particular NIC set to load as a module in the kernel. I can help you with kernel compilation if you have not tackled that before. It's not tough, so if you haven't done it don't worry. You can have both NICs up and that won't be a problem.

I myself really think it is either a patch cord or your NIC right now. I'm inclined to think the NIC. Again, if there were collisions, your switch would see them. It is possible however, the switch port is bad. Not likely. But possible. By moving to a new port we can isolate that. If we were in close proximity, I would come out and give you a hand. I have a Ethernet test set that will blast traffic at line speed and give a full report on everything. Odds of that are pretty slim obviously. I'm in the Pacific Northwest.

I ran the lsmod and found out that my NIC is a bcm5700 with an outdated driver. I applied the patch and reset the interface. I checked back about 2 hours and I see 52 collisions instead of hundreds and thousands in this case. I will do a swicharooney on the cable tomorrow. If I still see collisions on the intergrated NIC, I will install a new NIC on this server. I will also on the lookout for any collisions on the switch side as well.

Clark - Thanks for your responses and an offer to help out. Unforntunately, I live couple of thousand miles away.

I will post the outcome of this NIC issues.

Thanks,
JB

Good luck and your welcome. I would be suprised to find it was a driver...but I guess that is possible.