EIGRP flapping help?

[wormvine1]

Hey All,

I am having an issue that I cannot figure out.

I am currently having EIGRP neighbor adjacency flapping on multiple T1 data interfaces.

This issue started when I appplied a basic QOS to the serial T1 interfaces.

When the QOS policy is removed the Peer termination messages stop and the link has no issues.

I have tried multiple service policies to try and stop the issue.

Also the serial interfaces start accumulating errors when any of the listed policies are allowed

The router is a Cisco 2921.

I have 4-VWIC4-4MFT or somethings in the router.

Anybody have any ideas?  TIA

Here is my QOS:

class-map match-any VOIP
  description QOS MATCHING RULES
match ip dscp ef
match ip dscp af41 (this is our Avaya control traffic)

class-map match-any EIGRP

  match protocol eigrp

policy-map T1
class VOIP
  priority percent 50
class EIGRP
  bandwidth percent 5

policy-map VOIP50
description VOICE QOS
class VOIP
  priority percent 50
class class-default
  fair-queue

policy-map VOIP
description VOICE QOS
class VOIP
  priority percent 70
class class-default
  fair-queue

policy-map VOIP30
description VOICE QOS
class VOIP
  priority percent 30
class class-default
  fair-queue

[Julio Carvajal]

Hello Adam,

So we are getting peer termination logs on this side of the EIGRP neigbhorship, which means the other peer is sending a "Reset" packet for the EIGRP process.

The question would by why is it sending that information?

Based on the errors you are seeing on the interface counters,etc I think the other side is not receiving hello packets from this side on the right time (Hold-Time reached).

A debug will definetly let us know what's the cause of the reset.

EIGRP by default reserves 50% of the available bandwith for the Control-plane processing for EIGRP and in this case I can see you are providing 5 % (On the first one) which seems to be too low but anyway let's run the debug and provide us the output.

For more information about Core and Security Networking follow my website at http://laguiadelnetworking.com

Any question contact me at jcarvaja@laguiadelnetworking.com

Cheers,

Julio Carvajal Segura

[wormvine1]

Which debug should I capture?

I am not that experienced in running debugging.

I must say I am pretty amazed that EIGRP would reserve 50% of the bandwidth of a T1 link for itself.

Did I read that correctly?  

I looked at the hello and hold timers and they are 5 second Hello and 15 sec Hold.

That might not be enough for a busy T1 link. 

Thanks for your input!

Adam

[paul driver]

Helllo
have you tried removing.the specified eigrp classifcation from the class class-default as routing protocols are by default assigned to this class anyway -For on such a.link leaving the class class-default for igp should be applicable

res
paul

Sent from Cisco Technical Support Android App

[wormvine1]

Hello,

My initial QOS was just matching dscp EF and AF41 and giving it a priority of 30% or 50% or 70%..

I added the EIGRP class because the basic QOS policy, which is taken right off some Avaya documents caused the peer termination flapping. 

Thanks for the suggestions!

Adam

[JohnTylerPearce]

Assuming this a a full T1 and not a fractional T1, then your bps should be 1544000. So if you give the VoIP class "priority percent 50", if there is congestion on the link, then it's going to give 50% of the interface "bandwidth" to that queue, and that queue will be servicd untill it's empty, which will then let it go to other things (That is if congestion is taking place).

The above is correct about EIGRP taking up to 50% of bandwidth by default or "can" I should say. Try doing a 'show policy-map int x/x" and check out the values. You can also change the default bandwidth that EIGRP reserves.

Also, what is in 'show logging' when the neighbor relationships are peering. Are any interfaces going down?

[wormvine1]

Hey John,

Thanks for the info.

The T1 is a running at 1544 and not fractional.

Basically the T1 feeds a small lodge.  There are maybe 3-5 VOIP phones connecting to a very large Avaya system across the T1.  Maybe 2-4 Workstations.  I have 5 or 6 similar locations that do the same thing.

Oddly I checked the utilizaton on one link that was bouncing a lot and it was like 30/255 both in and out.

I even changed the hold timer to 60 seconds and it did not help.

If I follow what you are saying.   Say all 5 phones are in use at the same time and some user starts streaming a youtube video.  Since I gave the VOIP calls 50% and the youtube video took all the rest,  the EIGRP hellos are dropping or something?

I had to remove the policies becacause I have a SLA I have to maintain at these locations and I can't have links flapping all day.  My boss wants to just use static routes cause the locations only have one or 2 vlans on site.

If I ping the edge router I will lose pings during the flapping.  I did notice that the link did go down as well during the worst of it..

I will re-apply the qos on one of the less critical sites and see what else I can see happening.

I will run netflow and see the bandwidth utilization as well. 

Thanks for your suggestions!

Adam