On Thu, 2 Aug 2001, David Sinn wrote:
> Drops in the input queue are only for process switched packets (there is
> some exceptions, but given the next item it is pertinent). Since at the
> snap-shot provided we have 8 packets in the queue, you probably have
> something odd going on causing the packets to be punted to the CPU (and
> thus high CPU load).
>
> Are you missing a route?
> Are the packets going to a local destination that doesn't exist, and you
> are ARPing for the MAC address to send them too?
> Are the packets destined for the router?
> Do you have route-caching disabled on this interface, or the outgoing
> interface for the destination traffic?
>
> "ip route-cache same-interface" is only useful if you are on a 7500 with
> distributed switching turn on on VIP's. Since this is a 7100 all the
> processing is on the central CPU.
Fast caching is done on outbound interface, same-interface is the only way
to get the route cached regardless of CPU location.
-Scott
>
> David
>
>
> -----Original Message-----
> From: George Robbins [mailto:grr@shandakor.tharsis.com]
> Sent: Wednesday, August 01, 2001 7:07 AM
> To: cisco-nsp@puck.nether.net; jcv@vbc.net
> Subject: Re: [nsp] 7100 and small packets
>
>
> I'm not sure that the CPU loading is your worst problem.
>
> > Output queue 0/40, 0 drops; input queue 8/75, 1478 drops
>
> The load realated packet loss is .0015% though it will be worse
> at peak loading. Since this is input drops, it means packets
> coming in more quickly than old ones can be sent out.
>
> > 208760 input errors, 208760 CRC, 0 frame, 0 overrun, 0 ignored
> > 0 watchdog, 0 multicast
> > 103898 input packets with dribble condition detected
>
> These suggest that you have an ethernet problem, possibly wiring
> or perhaps a switch that's not 100% compatible as far as switching
> modes or flow control, or is having loading problems on it's own.
>
> In any case, you want to review the setting and error statistics
> for the other end of the link.
>
> As far as CPU utilization, if you're not using "ip route cache cef"
> plus "ip route-cache same-interface" do it.
>
> Generally if the router has lots of routes, as in an internet gateway,
> you need to be using CEF, if it has only a few such as an internal
> corporate network, then flow or normal (fast) cache will work about
> as well.
>
> George
>
> > Date: Wed, 1 Aug 2001 14:44:14 +0100 (BST)
> > From: Jean-Christophe Varaillon <jcv@vbc.net>
> > Subject: [nsp] 7100 and small packets
> >
> > Hi,
> >
> > On a 100 M Full-duplex interface of a Cisco 7100, I receive some very
> > small packet due to Games application and/or VOIP traffic.
> >
> > Because of this I have an abnormal CPU utilization rate.
> > Therefore, I have some packet loss when I run somes ping.
> >
> > kg6.uk0#sho int fa0/0
> > FastEthernet0/0 is up, line protocol is up
> > Hardware is DEC21140A, address is 0003.3184.8800 (bia
> 0003.3184.8800)
> > Internet address is 192.168.0.21/25
> > MTU 1500 bytes, BW 100000 Kbit, DLY 100 usec,
> > reliability 255/255, txload 11/255, rxload 13/255
> > Encapsulation ARPA, loopback not set
> > Keepalive set (10 sec)
> > Full-duplex, 100Mb/s, 100BaseTX/FX
> > 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 22:17:20
> > Queueing strategy: fifo
> > Output queue 0/40, 0 drops; input queue 8/75, 1478 drops
> > 5 minute input rate 5453000 bits/sec, 1975 packets/sec
> > 5 minute output rate 4342000 bits/sec, 1983 packets/sec
> > 95383572 packets input, 1399376380 bytes
> > Received 162215 broadcasts, 0 runts, 294 giants, 1478 throttles
> > 208760 input errors, 208760 CRC, 0 frame, 0 overrun, 0 ignored
> > 0 watchdog, 0 multicast
> > 103898 input packets with dribble condition detected
> > 95244094 packets output, 1975945187 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
> >
> >
> > I would like to know what alternative I have to solve this trouble.
> >
> > Regards,
> > Jean-Christophe.
> >
> >
> >
>
This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:12:48 EDT