Re: input/output Qeue Drops

kjbarrass · ‎01-09-2007

I have a 1851 router connected to a 4507 switch via a LES10 circuit. The users on the 1851 router say they get intermitent packet loss over the link. Looking at it there are output and input queue drops. Anyone know what mught be causing this as the circuit is monitored over 5 min averages and is peaking at about 5 meg.

The 4507 is in layer 2 mode only and has a 6509 hanging off it for routing.

The interface on the 4507 is an access port and both ends are set to 10/full duplex.

Show Interface details below

Any advise appreciated.

Cheers, Kev

Interface on Cisco 2851 Router connecting to Cisco 4507

sh int gi0/0

GigabitEthernet0/0 is up, line protocol is up

Hardware is MV96340 Ethernet, address is 0012.d974.77f8 (bia 0012.d974.77f8)

Description: Link to 4507

Internet address is *.*.*.*/29

MTU 1500 bytes, BW 10000 Kbit, DLY 1000 usec,

reliability 255/255, txload 1/255, rxload 5/255

Encapsulation ARPA, loopback not set

Keepalive set (10 sec)

Full-duplex, 10Mb/s, media type is T

output flow-control is XON, input flow-control is XON

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

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

Last clearing of "show interface" counters 1d00h

Input queue: 0/75/2/0 (size/max/drops/flushes); Total output drops: 2933

Queueing strategy: fifo

Output queue: 0/40 (size/max)

5 minute input rate 206000 bits/sec, 20 packets/sec

5 minute output rate 45000 bits/sec, 17 packets/sec

8333836 packets input, 1705553199 bytes, 2933 no buffer

Received 4381 broadcasts, 0 runts, 0 giants, 1 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

7237272 packets output, 935899431 bytes, 0 underruns

0 output errors, 0 collisions, 0 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

Interface on Cisco 4507 "in layer 2 mode only" connecting to Cisco 2851

GigabitEthernet7/8 is up, line protocol is up (connected)

Hardware is Gigabit Ethernet Port, address is 000f.2386.f657 (bia 000f.2386.f657)

Description: Link to 2851

MTU 1500 bytes, BW 10000 Kbit, DLY 10 usec,

reliability 255/255, txload 44/255, rxload 3/255

Encapsulation ARPA, loopback not set

Keepalive set (10 sec)

Full-duplex, 10Mb/s

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

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

Last input 00:00:38, output never, output hang never

Last clearing of "show interface" counters 23:37:51

Input queue: 0/2000/0/0 (size/max/drops/flushes); Total output drops: 1661

Queueing strategy: fifo

Output queue: 0/40 (size/max)

5 minute input rate 125000 bits/sec, 122 packets/sec

5 minute output rate 1738000 bits/sec, 180 packets/sec

7130424 packets input, 955610566 bytes, 0 no buffer

Received 1559 broadcasts (1559 multicast)

0 runts, 0 giants, 0 throttles

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

0 input packets with dribble condition detected

8274360 packets output, 10202823903 bytes, 0 underruns

0 output errors, 0 collisions, 0 interface resets

0 babbles, 0 late collision, 0 deferred

0 lost carrier, 0 no carrier

0 output buffer failures, 0 output buffers swapped out

ebanks2006 · ‎01-09-2007

It's very easy to overwhelm a 10Mbps link. If you try to shove too much data down that pipe, you're going to overflow the queue. My best guess is that in occasional bursts, the 10Mbps link is getting saturated, and your users are experiencing the packet loss.

You might consider using Generic Traffic Shaping (GTS) to throttle the user's bandwidth to 10Mbps before it has to traverse the LES10 circuit between the switches. GTS is pretty graceful in circumstances like this, and may reduce the packet loss.

ebanks2006 · ‎01-09-2007

Another thought - I was unfamiliar with LES10, but after reading up on it, I'm understanding it as the 10Mbps flavor an SHDS fiber connection; British Telecom likes to call it "LES10". Is it possible to get some circuit statistics from your provider? They may be able to offer some insight into the packet loss, if my theory turns out to be wrong.

kjbarrass · ‎01-10-2007

Thankyou for the above I will get in touch with our telco and see what they can offer us stats wise. Is there an IOS command to see what the peak/max unicast and multicast traffic flow was through an interface since last clear counters.

ebanks2006 · ‎01-10-2007

I am not aware of an IOS command that will give you the peak load since the last "clear counters". I did some poking around on a 3750 today, and came up dry. Most of what I found relating to finding traffic peaks had to do with historical analyzers. Most of these analyzers pull their data either via SNMP and/or by analyzing NetFlow data exported from the devices you've modeled.

If you are interested in some free and popular SNMP graphers, you could do worse than MRTG or Cacti. I'm sort of partial to Cacti, myself. Tidy little interface.

http://oss.oetiker.ch/mrtg/

http://cacti.net/

A commercial solution that would be a GREAT tool in your arsenal is the SolarWinds Engineer Edition (I think Windows only). It'll do a lot of wonderful things, including monitor bandwidth utilization, and graph it historically. It's powerful, and once you have it you'll wonder how you lived without it. But they don't give it away.

http://solarwinds.net/products/toolsets/engineer.aspx