RX Packet Errors - slow LAN xfers.

[NJC]

This is a recent problem, and I can't pinpoint any change/upgrade that would cause this. Rsync transfer from Client to Server:

Quote:
sent 11756196 bytes received 1032741 bytes 138258.78 bytes/sec
total size is 144333466390 speedup is 11285.81

Pinging back and forth from each machine is fine. No Ifconfig errors Client, but Server has RX packet errors.

Code:


eth0 Link encap:Ethernet HWaddr 00:11:25:37:ee:44

inet addr:192.168.1.101 Bcast:192.168.1.255 Mask:255.255.255.0

inet6 addr: fe80::211:25ff:fe37:ee44/64 Scope:Link

UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1

RX packets:41786 errors:2157 dropped:0 overruns:0 frame:2157

TX packets:34138 errors:0 dropped:0 overruns:0 carrier:0

collisions:0 txqueuelen:1000

RX bytes:55615449 (55.6 MB) TX bytes:4737538 (4.7 MB)

What do you think? 138258.78 bytes/sec is rubbish.

 

[OVERKILL]

Brand of NIC/NIC Chipset in each system? Switch? Are we dealing with a cheap or decent switch?

For example, sh int fa0/1 on my 2960 shows:

Code:


DFSWITCH#sh int fa0/1

FastEthernet0/1 is up, line protocol is up (connected)

Hardware is Fast Ethernet, address is 0024.133b.6301 (bia 0024.133b.6301)

Description: WAN

MTU 1500 bytes, BW 100000 Kbit, DLY 100 usec,

reliability 255/255, txload 1/255, rxload 1/255

Encapsulation ARPA, loopback not set

Keepalive set (10 sec)

Full-duplex, 100Mb/s, media type is 10/100BaseTX

input flow-control is off, output flow-control is unsupported

ARP type: ARPA, ARP Timeout 04:00:00

Last input never, output 00:00:00, output hang never

Last clearing of "show interface" counters never

Input queue: 0/75/0/0 (size/max/drops/flushes); Total output drops: 83

Queueing strategy: fifo

Output queue: 0/40 (size/max)

5 minute input rate 10000 bits/sec, 10 packets/sec

5 minute output rate 128000 bits/sec, 23 packets/sec

32764947 packets input, 36167332690 bytes, 0 no buffer

Received 1141 broadcasts (0 multicasts)

0 runts, 0 giants, 0 throttles

0 input errors, 0 CRC, 0 frame, 0 overrun, 0 ignored

0 watchdog, 0 multicast, 0 pause input

0 input packets with dribble condition detected

31495316 packets output, 11390218965 bytes, 0 underruns

0 output errors, 0 collisions, 1 interface resets

0 babbles, 0 late collision, 0 deferred

0 lost carrier, 0 no carrier, 0 PAUSE output

0 output buffer failures, 0 output buffers swapped out


Can you check the ports on the switch to see if they are showing errors?

 

[javacontour]

I think it's a hardware issue. The errors are frame errors. I'd start cheap by removing any other hardware from your network. I.E. remove any hosts not needed. Perhaps look at changing the ethernet cable and if that doesn't help, change ports on the switch.

2157 errors out of 41K frames is about a 5% error rate and will likely slow TCP as frames that are in error will need retransmission and TCP will slow down if frames containing TCP segments are lost and require retransmission.

So while 5% doesn't seem like much, it can cause problems like what you see on the network.

 

[NJC]

2 machines on LAN connected via old Linksys router - desktop (Client) and Server.

Client NIC:

Code:
Ethernet controller: Atheros Communications AR8121/AR8113/AR8114 Gigabit or Fast Ethernet (rev b0)

Subsystem: ASUSTeK Computer Inc. P5KPL-CM Motherboard

Flags: bus master, fast devsel, latency 0, IRQ 44

Memory at febc0000 (64-bit, non-prefetchable) [size=256K]

I/O ports at ec00

Capabilities: [40] Power Management version 2

Capabilities: [48] MSI: Enable+ Count=1/1 Maskable- 64bit+

Capabilities: [58] Express Endpoint, MSI 00

Capabilities: [100] Advanced Error Reporting

Kernel driver in use: ATL1E

Kernel modules: atl1e

Server:
Code:
Ethernet controller: Intel Corporation 82562EZ 10/100 Ethernet Controller (rev 02)

Subsystem: IBM Device 02c7

Flags: bus master, medium devsel, latency 66, IRQ 20

Memory at c0120000 (32-bit, non-prefetchable) [size=4K]

I/O ports at 2400

Capabilities: [dc] Power Management version 2

Kernel driver in use: e100

Kernel modules: e100

 

[OVERKILL]

Do you have another switch you could drop in to test with?

 

[NJC]

Negatory on different switch to try. My CAT5 ethernet is about 20' long to Server too, if that matters. But this is fairly recent problem, as I *used* to get better xfers with this hardware - at least in the MB's.

 

[javacontour]

How about different ports on the switch? Change the port used by the host experiencing the frame errors first and see if you get additional frame errors.

 

[javacontour]

The other thing to consider is that gigabit ethernet should auto-crossover, so you can take the switch out and cable the two hosts directly with a known good ethernet cable, taking the swtich out of the equation completely.

I.E. you don't need a crossover cable if one of the hosts has a gigabit ethernet interface. Even if you are not working at gigabit speeds, the interface should detect it's connected directly to another nic and do the cross over on the chip.

(I'm assuming the Atheros claiming to be Fast or Gigabit is actually a Gigabit interface.)

 

[Colt45ws]

Its likely the switch. I failed a Netgear 10/100 not too long ago. It showed up as errors on large packets. Just pinging a host would never show any errors, but forcing max size packets would cause about a 10% drop. It was enough to cause WinSCP to lockup and crash.

 

[NJC]

Any comments on these 3 wired routers - 2 Trendnets and a Netgear? I think the TRENDnet TW100-S4W1CA should suffice for my setup.

 

[javacontour]

So have you narrowed it down to the switch?

 

[NJC]

I haven't identified switch yet, just advance planning. But I'll drag the Server box up from crawl space and test it with another cable.

No luck in directly connecting the 2 devices either. Too bad.

 

[Rand]

what did you do to directly connect the devices. you have to manually set ip's etc

 

[NJC]

Originally Posted By: Rand
what did you do to directly connect the devices. you have to manually set ip's etc

How would I do that on Ubuntu machines? 192.168.1.100 = Client and 192.168.1.101 = Statically assigned Server. I plugged ethernet from NIC to NIC on both machines.

 

[OVERKILL]

You can't borrow a switch? LOL!

 

[NJC]

I can probably borrow one from our IT dept, but I'm always begging for spare parts. And laugh away, I'm not a network guy.

 

[OVERKILL]

That's OK, I have better equipment in my house for my tiny home LAN than most SMB networks have, LOL

If you need one, we've got a Victoria office, and I've got a spare cheap 8-port (no, it isn't Cisco) switch there that I could probably lend you to troubleshoot with.

Let me know.

-Chris

 

[simple_gifts]

If the switch has indicator lights, make sure it is also full duplex. That burned me once.

 

[NJC]

Thanks Chris for the offer. There's a fellow at work who can lend one, although not until he's back from Christmas Jan 10th. I can hobble along with KB/s until then.

simple_gifts, the Full light is on on front of router.

 

[NJC]

I just did a transfer from Client Nautilus to Server and it was 8MB/s - but a few days ago it was 400KB/s.

I don't know which of the fraggin' about made the difference? Rsync blazed along heartily @ "sent 1587740267 bytes received 615292 bytes 7474614.40 bytes/sec" too.