EIGRP link auto shut

[DPanganiban]

Hi all!

This is my first post here.

I need your expertise or recommendations on an EIGRP setup.

here's the scenario.

ROUTERA<--ROUTERB--->ROUTERC

Router B is primary forwarding traffic to ROUTER A.

Everything works fine.

My only problem is this. I have web applications used by the client who are affected most of the time when the link goes down.

When ROUTER B's link to A goes down, it will route to ROUTER C which is good. however users of the web application need to  reconnect to their apps.

when ROUTER B's link to A goes up, user need to reconnect again. The sites are session based.

So the question is. is there a way to shut down the interface automatically so that it will pass traffic via Router C?

cause Everytime there is a network flap, the network re convergences so fast and the routes keep on updating.. this is good of course but not to my case as for the web apps user.

any information is  appreciated.

thanks!!!

Dp

[andrew.prince]

Dip,

Question - are the websites they connect to external?

[DPanganiban]

Hi Andrew,

The websites are from internal. they were redistributed via BGP.

I am thinking of HSRP but that would mean that  instead of going to Router A, I will make 2 connections pointing to Router C instead, removing the link to A.

The other way is to play with the hold/hello timers but it would be nice for the Routers to determine the link that went down is up and is really stable..

thanks for your time.

dip

[andrew.prince]

Dip,

Please expand on "The websites are from internal. they were redistributed via BGP."  How do the remote subnets get into the EIGRP routing table?  How do the clients know the default gateway?

[DPanganiban]

Hi!

The websites were internal sites from the client redistributed via EIGRP/[ex]/BGP.

Their default gateways were automatically assigned via Vlan.

However, when the links shift to the failover, the users tend to lose their sessions and therefore need to login again in the website mentioned.

HSRP might solve this.

I haven't tried doing HSRP from 1 location going to 2 different locations but I know its possible and there is the question of having 2 different ISPs.

A switch in the middle might do a trick.

Thanks so much for replying to this post.

Dip

[andrew.prince]

Using a dynamic next hop protocol to fix a routing issue is not the answer.

The issue is EIGRP dead peer detection, and route sucessor implementation.

Change the hello & hold timers on the links that connect the EIGRP peers together.

EIGRP default hello is 5 seconds, hold is 15 seconds.  The users TCP sessions will timeout well before the hold time is reached.

Change the hello to 1 second and hold to 3 seconds.

[DPanganiban]

Hi Andrew,

Appreciate the response.

I think i have to try those values as it is also recommended by Cisco. I didn't configure the devices but upon checking, it is not standard. the

i forgot to mention.. the EIGRP runs inside a GRE tunnel.. recently we noticed one server maxxed out the bandwidth causing the tunnel to flap.

the only successor went online.. however everytime the tunnel goes back up.. it would choose it again over the failover..

i forgot to mention about the tunnel.. sorry.. the reason why EIGRP is placed inside the tunnel is due to physical link issues.

we noticed when the link goes down and when it goes up, we had to reboot the Router.. so to eliminate this, one suggested the tunnel.

i  heard GRE tunnels have limitations.. i noticed it caps bandwidth to 8MB?

il discuss the values with our team as it is not standard and try those values inside the tunnel.

the values for both end [tunnels]should be the same right?

tunnel configs for primary.

[this is the config of ROUTER B and ROUTER A] ROUTER C is the failover device.

setup is : ROUTER A-->tunnel<--ROUTER B-->FE<--CORESWITCH<--FE-->ROUTER C.

router 1:
ip hello-interval eigrp 10 1
ip hold-time eigrp 10 2
ip tcp adjust-mss 1400
keepalive 1 4

router 2:
ip hello-interval eigrp 10 1
ip hold-time eigrp 10 2
ip tcp adjust-mss 1400
keepalive 1 3

CORESWITCH

:

ip hello-interval eigrp 10 1
ip hold-time eigrp 10 2
keepalive 1

thanks so much!!

dip[

[andrew.prince]

GRE tunnels by default will remain up - as aslong as a route/interface for the destination is in the routing table, even if there is a break in the link mid point.  You can configure a GRE tunnel on a router with NO interfaces actually connecrted to anything and the tunnel will be up/up.  To overcome this configure tunnel keepalives I suggest to match the EIGRP keepalives 1 hello dead/hold 3

If there is a break in the circuit connectivity - the tunnel will drop, the EIGRP neighbour will be lost and the sucessor route will take over.

I have 60+ GRE tunnels, all running over Public internet - all constantly transmiting 15mbs and over.  You will find the reason for the cap will be the access circuit speed.  Check to see what the circuit speeds are - this will be your bottleneck.

[DPanganiban]

Hi Andrew,

While i was updating my post.. i got a notification of your reply.

that's interesting.. and yeah i saw that before that it comes up/up without any interfaces connected.

i didn't know that you can put in on production and that make sense.

currently the tunnels have its Tunnel Source. so safe to remove the source and destination  as long as EIGRP configs with the right hello/hold timer and of course the routing table are there?

[andrew.prince]

If you remove the tunnel source & destination the tunnel will go down - the point I was making is that without keepalives the tunnel

will remain up/up with no network connectivity, if the routes are valid in the routing table.

Apply the EIGRP Hello/Hold timers & the Tunnel Keepalives - do not remove any tunnel config.