cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
25799
Views
0
Helpful
14
Replies

Seeing What Prefixes EIGRP Is Advertising

mfarrenkopf
Level 1
Level 1

Is there a command available on the 6500 that I can use to see what prefixes it is advertising directly to a neighbor?

The diagram is detailed and complex, but the simplest problem statement is that it doesn't look like my 6500 distribution switches are advertising certain prefixes to one of the 6500 access switches.  I don't know whether this is an issue of the distribution switches not sending the prefixes down to the access layer (they should be; the route originates on a different set of access switches) or my access switch is dropping the prefixes.  I don't see them in the topology table at all.

If the prefix isn't being advertised, I need to troubleshoot the distribution.  If it is and it's being ignored, I need to troubleshoot the access.  There are no obvious conditions that would prevent the access switch from getting the prefixes -- interfaces aren't passive, no distribute lists at work, everthing in the same AS, I have neighbor relationships (and I am getting other prefixes over these links, and these prefixes are being advertised to other access switches), auto-summarization is off, split horizon is still on . . .

I'm stumped.  I have a very detailed diagram of all of the metrics and links and I don't see any reason why my access switch shouldn't be getting the prefixes.

6509 chassis, dual sup 720 3B, 12.2(33)SXI4a advanced enterprise services IOS.

1 Accepted Solution

Accepted Solutions

dslice
Level 1
Level 1

You can use the event log to see this but it's really a lot of data to slog thru. First of all, enable all three event types and increase the size of the log.

Router EIGRP 1

EIGRP log-event-type dual xmit transport

EIGRP event-log-size 10000

Clear the log by doing:

Clear ip EIGRP events

Clear the neighbor or do something to cause the routes to be seen on the peer, then display it:

Show ip EIGRP events

Most recent events on top

Sent from Cisco Technical Support iPad App

View solution in original post

14 Replies 14

Reza Sharifi
Hall of Fame
Hall of Fame

I think, the closest command is this:

Since you already know what prefixes your distro switch is advertising to the access switch, try this command on the access switch

Switch#sh ip eigrp 1 topology ?

  A.B.C.D          Network to display information about

  A.B.C.D/nn       Prefix /, e.g., 192.168.0.0/16

  active           Show only active entries

  all-links        Show all links in topology table

  detail-links     Show all links in topology table

  pending          Show only entries pending transmission

  summary          Show a summary of the topology table

  zero-successors  Show only zero successor entries

  |                Output modifiers

 

Switch#

add A.B.C.D at the end and see if you get the same routes from distro switch

HTH

Hi Reza,

Thanks for the reply.  Yes, I've used the show ip eigrp topology command many times in troubleshooting this, and the prefix just doesn't appear at all from the distribution.

This is a layer 3 routed design (no VLANs spanned from the distribution).  There are other 6500s and 3750s connected to the distribution, but they are getting the T2 prefixes as they should.

* T2, which is originating the prefix, has routed gig uplinks to D1 and D2.

* D1 and D2 have an interregion routed gig link.

* F1 and F2 have dual routed gig uplinks to D1 and D2.

* N1 and N2 are Nexus 5548s that were just installed that have routed 10 gig uplinks to F1 and F2.  They also have an interregion between themselves.

T2 is advertising the prefixes to D1 and D2.  D1 and D2 are advertising the prefixes between each other and to F2.  F2 is advertising the prefixes to N1 and N2.  N1 and N2 are advertising the prefixes to F1.

Looking at the metrics, D1 and D2 should be advertising the routes to F1.  They are advertising other prefixes to F1.  T2's prefixes are the only prefixes that F1 is getting through N1 and N2.  If I turn off the links between F1 and N1/N2, F1 doesn't have the prefixes at all.  By sheer coincidence, the traffic ends up taking the correct links to D1 and D2 because of the default route.

Neighbor relationships are in place and functional, as evidenced by the fact that F1 is getting other prefixes from D1 and D2.  Again, there are no distribute lists, route maps, passive interfaces, redistribution, no variance commands, neither F1 nor F2 is stubbed, N1 and N2 have delays on their 10 gig links to make those routes less palatable.  The routes that F1 is using are very much not preferred; the FD F1 is using is 8448, whereas if D1 and D2 were advertising to F1 they would have an AD of 3072, which would make F1's FD about 3328 (educated guess based on what's happening on other links).  And this is ONLY happening with T2's prefixes.  Using some careful output filtering, it was quickly apparent that this was going on with three prefixes and they all originate from T2.  But other access switches are getting the prefixes just fine.

It just flat out looks like EIGRP is deciding not to advertise the prefixes to F1 for reasons completely unknown to me.

I'm sure there's a debug command I could use, but this is our data center distribution, and I tend toward being very protective of my distribution switches when it comes to debug commands.  So that's why I was hoping there was a command that I was missing that would just tell me what it was advertising to a specific neighbor.

Thank you for taking the time to reply!

Hi,

Can you try this command "clear ip eigrp 1 neighbors x.x.x.x"

If above does not help:

Is there any way for you to delete the EIGRP process on F1 completely and reapply it again?

it would be nice if EIGRP had a show command just like BGP

sh ip bgp neighbors x.x.x.x received prefix-filter

Also, is there way for you to put a diagram together and  post it here, so I can follow your write up?

Thanks,

I tried clearing the neighbor relationships earlier today, both through a soft reconfig and a complete reset of the neighbor.  Didn't work.  Tried it from both F1 and D1 (didn't try it on D2 though) without success.

Right now, everything is hand drawn.  I'm happy to draw up a diagram, but it will likely be Tuesday before I get that done.  (I'm actually home right now but this was just driving me bats all day.)

Thanks!

Ok, I would be curious to know if deleting the process on Tuesday makes any difference.  On your first post, you noted:

6509 chassis, dual sup 720 3B, 12.2(33)SXI4a advanced enterprise services IOS.

I know that SXI4a is a pretty good image compared to earlier releases.  Are your sups in redundant SSO mode?

Thanks,

Reza

To completely clear and restart the process, I'll have to schedule a work in area.  I won't be able to do that before Tuesday.

Sup 720 3B in all 6500s.  D1, D2, F1, and F2 are all dual sup running in SSO mode.  T2 is a single sup that will soon be dual sup.

We just upgraded the data center to SXI4a about mid-October.  We'd been running it on our cores and other building distributions with no problems.  This is the first incident that is really making me scratch my head.

Parenthetically, I'd really like something that could do a campus-wide analysis of our routing and track routing changes.  There isn't much out there.  Packet Design's Route Explorer can neighbor with EIGRP, but it's the only product I've found so far, and it's over $100K.  If you have any other suggestions, I'm all ears.  I've got a programming background.  I'd love to write something on my own (and I'm toying with the idea a little bit).  But something a little less expensive that was available right now . . . that does EIGRP . . . could be very useful.

Look into eem scripting it will monitor routes for free if you can script a little

Matt

Sent from Cisco Technical Support iPad App

and did you capture the event log as I suggested when you did the clear neighbors to see what the sender was doing?

Sent from Cisco Technical Support iPad App

I tried resetting the neighbor relationships prior to posting here.

My organization gives us Monday off for the new year,  so it'll be Tuesday before I can try it.  But I will try it and I'll let  you know.

Actually, it seems that T2 IS using 10.76.255.16 for its router ID:

EIGRP-IPv4 Topology Table for AS(12)/ID(10.76.255.16)

Codes: P - Passive, A - Active, U - Update, Q - Query, R - Reply,

       r - reply Status, s - sia Status

Not sure how it happened, but now it's something I need to fix.  And it's not going to be easy.

Thank you for all of your help!  Your suggestion is what led me to gather the information I needed, so I'll mark yours as the correct answer!

Take care,

Matt

Hi Donald,

Thank you for the suggestion.  I had to up the log size to 200000, but it showed up:  I'm getting "Ignored route, dup routerid int: 10.76.255.16" on reception by F1.  We don't do manual router ID setting.  The relevant portion is:

113939 07:22:10.541 Ignored route, metric: 10.76.60.128/25 3328

113940 07:22:10.541 Ignored route, dup routerid int: 10.76.255.16

113941 07:22:10.541 Ignored route, metric: 10.76.61.0/27 3328

113942 07:22:10.541 Ignored route, dup routerid int: 10.76.255.16

113943 07:22:10.541 Ignored route, metric: 10.76.38.0/26 3328

113944 07:22:10.541 Ignored route, dup routerid int: 10.76.255.16

113945 07:22:10.541 Ignored route, metric: 10.76.255.18/32 131072

113946 07:22:10.541 Ignored route, dup routerid int: 10.76.255.16

I've confirmed that T2 does NOT have this router ID.  I did a grep through our config archive and there are no other routers that have this IP address.  This address is assigned to F1 and no one else.  We do have some VRF Lite running but we have not duplicated address space.  In fact, for each VRF we assign its own independent loopback, so it shouldn't even be trying to use the wrong loopback across routing tables.

Thoughts on where else to look?  T2's router ID should be 10.76.255.18.

Thank you,

Matt

dslice
Level 1
Level 1

You can use the event log to see this but it's really a lot of data to slog thru. First of all, enable all three event types and increase the size of the log.

Router EIGRP 1

EIGRP log-event-type dual xmit transport

EIGRP event-log-size 10000

Clear the log by doing:

Clear ip EIGRP events

Clear the neighbor or do something to cause the routes to be seen on the peer, then display it:

Show ip EIGRP events

Most recent events on top

Sent from Cisco Technical Support iPad App

I have no problem slogging.  I have a problem with a network that doesn't seem to be advertising routes as it should!

Thank you for the reply.  Not exactly what I was looking for (I was hoping for a convenient command that would show me the prefixes it had advertised), but this should work.  Thank you very much for your input.  I will try this on Tuesday and report back.

Hi,

Can you check the output of "show ip eigrp interface brief" to find out if the eigrp is enabled on the interface? I'm assuming that prefix is configured on one of the interface of your distribution switch.

also check the output of show ip eigrp topology to check if the prefix is listed on topology table.

also, if can enable "debug ip eigrp event" swtiches, this should give you the list of network that the distribution switch advertise to its neighbor.

also you can create an access list filtering your interest area and apply this ACL along with the debug command to minimize the no of output seen from the "debug ip eigrp event" (for that matter any debug command) to troubleshoot it further.

Hope this helps :-)

-Vijay

Getting Started

Find answers to your questions by entering keywords or phrases in the Search bar above. New here? Use these resources to familiarize yourself with the community:

Review Cisco Networking products for a $25 gift card