Network problems, bad card?

Over the past few weeks, I have noticed my network speeds dropping on my LAN, specificially copying a file to/from the server would start out at the expected rate of a few megs a second, but soon after would drop into kb/second and even down to 0 before picking back up. I did some copying between two machines on the same switch to determain that the issue is just with the server. I've changed the network cable and port, so now I am looking at the card. ifconfig reports this:

eth0 Link encap:Ethernet HWaddr 00:50:DA:XX:XX:XX
inet addr:192.168.0.1 Bcast:192.168.0.255 Mask:255.255.255.0
inet6 addr: fe80::250:daff:fe0e:8e/10 Scope:Link
inet6 addr: fe80::50:da0e:8e/10 Scope:Link
EtherTalk Phase 2 addr:65280/85
UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
RX packets:4855974 errors:2 dropped:0 overruns:1004 frame:2
TX packets:5299964 errors:0 dropped:0 overruns:0 carrier:137
collisions:0
RX bytes:4233389688 (4037.2 Mb) TX bytes:4262115146 (4064.6 Mb)

eth1 Link encap:Ethernet HWaddr 00:10:4B:XX:XX:XX
inet addr:68.64.8.66 Bcast:68.64.8.255 Mask:255.255.255.0
inet6 addr: fe80::210:4bff:fe74:d486/10 Scope:Link
inet6 addr: fe80::10:4b74:d486/10 Scope:Link
UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
RX packets:8108281 errors:0 dropped:0 overruns:89 frame:0
TX packets:4789725 errors:0 dropped:0 overruns:0 carrier:0
collisions:35187
RX bytes:3984511176 (3799.9 Mb) TX bytes:1897170285 (1809.2 Mb)

eth2 Link encap:Ethernet HWaddr 02:60:8C:XX:XX:XX
inet addr:192.168.1.1 Bcast:192.168.1.255 Mask:255.255.255.0
inet6 addr: fe80::260:8cdc:2422/10 Scope:Link
inet6 addr: fe80::60:8cff:fedc:2422/10 Scope:Link
UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
RX packets:0 errors:0 dropped:0 overruns:0 frame:0
TX packets:2331 errors:0 dropped:0 overruns:0 carrier:0
collisions:0
RX bytes:0 (0.0 b) TX bytes:448893 (438.3 Kb)

lo Link encap:Local Loopback
inet addr:127.0.0.1 Mask:255.0.0.0
inet6 addr: ::1/128 Scope:Host
EtherTalk Phase 2 addr:0/0
UP LOOPBACK RUNNING MTU:16436 Metric:1
RX packets:144366 errors:0 dropped:0 overruns:0 frame:0
TX packets:144366 errors:0 dropped:0 overruns:0 carrier:0
collisions:0
RX bytes:41553216 (39.6 Mb) TX bytes:41553216 (39.6 Mb)

eth0 is the internal LAN, eth1 is a direct connect to the cable modem and eth2 is another LAN connection for when the empeg is downstairs. The uptime when this was taken was 6 days. I am assuming the errors reported on the card could be causing this. Anyone have any insight before I play around with the network config of the box?

I experienced the same thing a while ago when I upgraded an office from an old 10bT hub to a 100bT switch. What should be been quite an improvement would slowly bring the network to a crawl. (WIth three computers, nothing major) I tried returning the linksys switch for a new one, and it did the exact same thing. Chaning to a different brand fixed the problem. I never figured out what it was that caused the slow down, but once the problem was solved I moved on and didn't worry about it.

Matthew

I would suggest that this might be Drakino's problem, as well. The fact that you have errors (which should never occur in a properly functioning network) only on the RX side of that one interface would tend to indicate that the device that that interface is connected to isn't functioning properly. You might want to try forcing 100Mbps/FD on both that device and on your interface and see if that clears things up.

Of course, this is just a guess.

I've seen this issue with Actiontek NIC's with Linksys and DLink switches before. I've also seen i with Netgear Switches and Linksys NICS. It seems to be a strange conflict of chipsets used. Switch the nic and it fixes it. Switch the Switch and it fixes it.

Tought me to stop buying cheap equipment :P

Yeah, i'm a netgear fan and don't really consider them cheap equipment, even if I should. I've never had any problems, and all of the hubs and switches have real metal cases instead of rounded plastic things with feet and the like... And blue PCB's look better than green...

Matthew

Well, NetGear is made by Bay Networks, which is a real networking company, as opposed to whatever random Taiwanese house makes D-Link, etc. So I think that you're right to hold them in slightly higher esteem (even if it is just due to the metal cases ).

Yeah, I had a flaky D-Link switch once. It got replaced with a far better Netgear Switch.

I can't force the switch to that speed, but I have tried forcing the NIC to 100fd with no change. I am still siding that the problem is the NIC, due to transfers working fine between other machines on the network. I've switched ports on the switch, and the issue always follows the server. I am probably going to swap the cable modem to eth2 and move the internal lan to eth1 as a test. (eth0 and 1 are both PCI based 3Com 905 cards, eth2 is a slow ISA 3Com card).

Well, a few days later after moving the LAN to the other 3c905 card, and the same problem exist. I'm begining to suspect the linux driver (never did like it much, Mark, you want to have a try at it? ). So next I'm going to toss in a NetGear card that is currently in the media system. The 3Com should work fine in the media system with it running Win2k.

Oh, and I did some more tests on the problem it's self. I sent a file to the server, and pinged it in another window. Pings appeared fine at first, less then 10ms. Then it started spiking to 60 or 70ish. It went back down, then shot up to over 600, and stopped responding to pings for 10 seconds. It then came back, and repeated this sequence. Checking ifconfig, the rx overrun seemed to increment a bit when this happened.

More tests over the network with other equipment seems to confirm the problem is limited to my server. If all else fails, I think a friend of mine has a 5 port 10/100 switch I can borrow to see if it impacts the problem at all.

I hate network problems like this. I prefer the ones where it stops working period.