EIGRP and SLA up and down

[ohforce55]

Hi,

We have two ASR routers in site A and other two ASR routers in Site B.

The routers in both sites are connected with P2P. (Using two different service providers and fibers, 600M each).

Last Friday, both connections went up and down. (From site A-1 to Site B-1 and from Site A-2 to Site B-2).

And Monday, something a little different happened on both routers...

G0/0/2 is the port that is connected to the circuit (between Site A router and Site B router).

SiteA-ASR-1# (same as SiteA-ASR2)

Jul 28 13:32:23.062 EDT: %DUAL-5-NBRCHANGE: EIGRP-IPv4 4907: Neighbor 10.50.31.x2 (GigabitEthernet0/0/2) is down: Interface PEER-TERMINATION received
Jul 28 13:35:28.048 EDT: %DUAL-5-NBRCHANGE: EIGRP-IPv4 4907: Neighbor 10.50.31.x2 (GigabitEthernet0/0/2) is up: new adjacency
Jul 28 17:34:15.678 EDT: %DUAL-5-NBRCHANGE: EIGRP-IPv4 4907: Neighbor 10.50.31.x2 (GigabitEthernet0/0/2) is down: Interface PEER-TERMINATION received
Jul 28 17:37:20.683 EDT: %DUAL-5-NBRCHANGE: EIGRP-IPv4 4907: Neighbor 10.50.31.x2 (GigabitEthernet0/0/2) is up: new adjacency

Another up and down on Monday.

Aug 1 12:36:55.131 EDT: %TRACK-6-STATE: 90 ip sla 90 reachability Up -> Down
Aug 1 12:40:05.133 EDT: %TRACK-6-STATE: 90 ip sla 90 reachability Down -> Up
Aug 1 14:05:56.111 EDT: %TRACK-6-STATE: 90 ip sla 90 reachability Up -> Down
Aug 1 14:09:01.111 EDT: %TRACK-6-STATE: 90 ip sla 90 reachability Down -> Up

SiteB-ASR-1#  (same as SiteB-ASR2)

Jul 28 13:32:23.043 EDT: %DUAL-5-NBRCHANGE: EIGRP-IPv4 4907: Neighbor 10.50.31.x1 (GigabitEthernet0/0/2) is down: interface passive
Jul 28 13:32:23.093 EDT: %SYS-5-CONFIG_I: Configured from console by EEMScriptMgr on vty0 (EEM:Passive_The_Link)
Jul 28 13:35:27.743 EDT: %TRACK-6-STATE: 90 ip sla 90 reachability Down -> Up
Jul 28 13:35:28.057 EDT: %DUAL-5-NBRCHANGE: EIGRP-IPv4 4907: Neighbor 10.50.31.x1 (GigabitEthernet0/0/2) is up: new adjacency
Jul 28 13:35:28.094 EDT: %SYS-5-CONFIG_I: Configured from console by EEMScriptMgr on vty0 (EEM:UnPassive_The_Link)
Jul 28 17:34:15.383 EDT: %TRACK-6-STATE: 90 ip sla 90 reachability Up -> Down
Jul 28 17:34:15.664 EDT: %DUAL-5-NBRCHANGE: EIGRP-IPv4 4907: Neighbor 10.50.31.x1 (GigabitEthernet0/0/2) is down: interface passive
Jul 28 17:34:15.734 EDT: %SYS-5-CONFIG_I: Configured from console by EEMScriptMgr on vty0 (EEM:Passive_The_Link)
Jul 28 17:37:20.382 EDT: %TRACK-6-STATE: 90 ip sla 90 reachability Down -> Up
Jul 28 17:37:20.697 EDT: %DUAL-5-NBRCHANGE: EIGRP-IPv4 4907: Neighbor 10.50.31.x1 (GigabitEthernet0/0/2) is up: new adjacency

Another up and down on Monday.

Aug 2 10:43:52.450 EDT: %DUAL-5-NBRCHANGE: EIGRP-IPv4 4907: Neighbor 10.50.31.17 (GigabitEthernet0/0/2) is down: BFD peer down notified
Aug 2 10:43:57.315 EDT: %DUAL-5-NBRCHANGE: EIGRP-IPv4 4907: Neighbor 10.50.31.17 (GigabitEthernet0/0/2) is up: new adjacency

I checked all the EIGRP config and couldn't really find the issue.

But, after my manager made some change on both routers just for Site A, it seems to be stable now, and he will apply that command for Site B later.

Before:

policy-map qos-GigabitEthernet-0-0-2
class replication
shape average 500000000 5000000
bandwidth remaining ratio 100
queue-limit 4096 packets
class class-default
shape average 100000000 1000000
bandwidth remaining ratio 100
service-policy standard-policy

router eigrp 4907
no bfd interface GigabitEthernet0/0/2

interface GigabitEthernet0/0/2
no bfd template To-NC  (and To-CA for other side router)

no bfd-template single-hop To-NC (and To-CA for other side router)

After:

policy-map qos-GigabitEthernet-0-0-2
class replication
shape average 480000000 1920000 0 account user-defined 24
bandwidth remaining ratio 100
queue-limit 4096 packets
class class-default
shape average 120000000 480000 0 account user-defined 24
bandwidth remaining ratio 100
service-policy standard-policy

bfd-template single-hop To-NC (and To-CA for other side router)
interval min-tx 300 min-rx 300 multiplier 3
dampening 3 1500 3000 5
echo

interface GigabitEthernet0/0/2
bfd template To-NC (and To-CA for other side router)

router eigrp 4907
bfd interface GigabitEthernet0/0/2

My questions is that

Although it is 600M, why the config says shape average 500000000? Should it be 600000000?

And why did he change it to 480000000 from 500000000?

Not only that but I will very appreciate what he did change and how it was resolved the issue.

Thank you!

[Julio E. Moisa]

Hi

Your manager has applied traffic shaping on the link, usually EIGRP uses up to 50% of the link utilization, probably he is just allowing 500M for the class replication. The rest will use the other 100Mb (class-default)

Now the question is, what is Class replication?

[ohforce55]

Julio,

Thank you again! :)

Sorry, I forgot to put the config for that class.

class-map match-any replication
match access-group name replication-out

ip access-list extended replication-out
permit ip 10.18.10.111 0.0.0.1 10.16.10.111 0.0.0.1

10.18.10.0(NC) and 10.16.10.0(CA) are  Network Server such as PDC,BDC,WINS,Mail.

They replicate each other thru the p2p circuit.

So when you said EIGRP usually uses up to 50% of the link utilization, if we have 600M, it uses 300M? But my manager forced it to allow 500M to be used out of 600M?

And  can you tell me what other 100M specifically is used?

[Julio E. Moisa]

Hi

You are welcome my friend, Just to clarify by default EIGRP will limit itself to use no more than 50% of the bandwidth on an interface to avoid EIGRP update packet lost, it can be reconfigurable, the limitation is applied to the EIGRP update packets only. Now it will not be affected for the real bandwidth applied on the link, so you dont need be worried about that. 

Now may I know to know why he is applying 500Mb to that class? Im not sure about the use of this link, 600Mb is a lot of bandwidth for the communication of these hosts.

The 100Mb or after the change 120Mb, are used for the rest of the traffic, probably the link supports up to 600Mb but Im not sure how the service was contracted. 

class class-default
shape average 120000000 480000 0 account user-defined 24
bandwidth remaining ratio 100
service-policy standard-policy

The class-default is used for the traffic who has not being applied to other class categories.  So in few words, if the traffic is not marked or included on the Replication Class will be passing through this default class. 

:-)

[ohforce55]

Thank you again and again.

I'm sorry to ask you another question but,

So how did that resolve the up and down issue?

[Predrag Jovic · Accepted solution]

no bfd command - removed reason for the second flap of interface and I guess template was changed later to be replaced with more relaxed timers.

Result is that now interface will be more resilient, but EIGRP will have longer reconvergence time in the case that directly connected interface is still in up up state and neighbor stops responding. For more details, please read article - Bidirectional Forwarding Detection Configuration

[Julio E. Moisa · Accepted solution]

Additional, I think he detected the traffic was reaching the capacity of the link, and EIGRP Updates packets could had been compromised, that could be a reason to have EIGRP Flapping.  

Shape unlike Police works with queue, so it will be retaining packets and then forwarding them to avoid exceed the limit applied.

[Predrag Jovic]

Check SiteB-ASR-1#  (same as SiteB-ASR2) configuration. Someone is running EEM script on that device.

Jul 28 13:32:23.043 EDT: %DUAL-5-NBRCHANGE: EIGRP-IPv4 4907: Neighbor 10.50.31.x1 (GigabitEthernet0/0/2) is down: interface passive
Jul 28 13:32:23.093 EDT: %SYS-5-CONFIG_I: Configured from console by EEMScriptMgr on vty0 (EEM:Passive_The_Link)

EEM script Passive_The_Link configured interface Gi0/0/2 as passive interface

Jul 28 13:35:27.743 EDT: %TRACK-6-STATE: 90 ip sla 90 reachability Down -> Up
Jul 28 13:35:28.057 EDT: %DUAL-5-NBRCHANGE: EIGRP-IPv4 4907: Neighbor 10.50.31.x1 (GigabitEthernet0/0/2) is up: new adjacency
Jul 28 13:35:28.094 EDT: %SYS-5-CONFIG_I: Configured from console by EEMScriptMgr on vty0 (EEM:UnPassive_The_Link)

EEM script UnPassive_The_Link configured interface Gi0/0/2 as non passive interface

Aug 2 10:43:52.450 EDT: %DUAL-5-NBRCHANGE: EIGRP-IPv4 4907: Neighbor 10.50.31.17 (GigabitEthernet0/0/2) is down: BFD peer down notified

BFD did not have timely answer from BFD responder and interface status was changed to down.

[ohforce55]

Than you for your answers!