Take the 2-minute tour ×
Server Fault is a question and answer site for system and network administrators. It's 100% free, no registration required.

I've got a Cisco 2811 (AdvIPServices 12.4(25e)) that is showing input errors on both its interfaces, all the errors are of the ignored type.

The counter for ignore errors is not increasing continuously, which made me think it would be happening in line with traffic bursts, but there really isn't a large amount of traffic running through it at any time.

Both the router interfaces are connected to a 4900 (IPBase 12.2(54)) series switch.

The only things I can see are:

  • There is a slight mismatch in duplex/speed negotiation settings, one side is set to auto/auto and one full/auto, however they have negotiated fine.
  • There have been a few fallbacks on the interface buffers, and some trims in the public buffers.

Any suggestions for things to look at? I will be correcting the duplex neg settings, and might try a buffer tune, but I don't think it will help.

share|improve this question

1 Answer 1

Both Ethernet ports should be running auto/auto, as should the 4900's ports. This isn't the source of an ignored packet, though - media mismatches will show up as CRC errors, input errors, etc..

How many ignored packets are you seeing and over what period of time? A few hundred over a period of months isn't that big of a deal - especially if you're not noticing a performance hits.

Are you measuring CPU usage at all? Even tracking on 'sh proc cpu history' might give some clues as to whether your counter increases correlate with high cpu usage (..which can contribute in its own right to these kinds of issues). Also confirm that cef is enabled on the box globally as well as on the individual interfaces. Run 'show int st' to get a sense of the ratio of process switched to fast switched packets - the number should be predominantly fast switched. If not, a particular feature or configuration is likely hanging things up. Are you doing NBAR, fancy firewall/IDS or lots of un-accelerated crypto?

If you're at all memory constrained I would go slowly with the buffer changes - going too far can cause more harm than good.

share|improve this answer
    
represents about 1% or less of the traffic. but not any NBAR or fancy firewall/IDS, a small amount of traffic is crypto. 'sh proc cpu history' has shown some spikes up near 90%, averaged out that could mean that the CPU is maxed out. could this cause input ignores? and if so wouldn't it cause outbound problems also? thanks for the show in st command, i'll check that out and check the ratio. –  elbekay Jun 5 '12 at 11:29
    
If high CPU is caused by process switching then problems can manifest as buffer issues (...which can then turn into ignores). I'd expect input to get hit before output for this kind of issue while output will get hit when output bandwidth rises to contention level. Also take a look at 'sh proc cpu' to see if one process is pulling a lot more CPU than the rest (specifically "IP Input" or similar). Also - any other interfaces in the box besides the two FE's? –  rnxrx Jun 8 '12 at 4:49

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.