convergence time- OSPF/BGP (MPLS)

[foysol_bgd]

Hi experts, 

Need some advice about OSPF and BGP convergence in a MPLS network. I am designing a small networks with 12 routers to provide L2/L3 connectivity to clients. 

Two routers will act as a BGP RR and will provide PE functionality as well at the same time.

Rest of the 10 routers will be dedicated PE routers.

Everything is working fine as expected. This is a simple setup with LDP, BGP and OSPF as IGP. 

I am planning to use BFD for OSPF quick convergence. 

"bfd interval 250 min_rx 250 multiplier 4"

and I have tuned the SPF timers as below 

 timers throttle spf 200 5000 5000

I am planning to keep the BGP timers as default for stability. But with client (EBGP) I will keep the timers low in future for quick detection. Let me know please what is your thought on this. 

[Giuseppe Larosa]

Hello foysol_bgd,

it looks like a good configuration you have already implemented it?

Hope to help

Giuseppe

[Jon Marshall]

Giuseppe !!!

Haven't seen you around for a long time.

Does this mean you will be posting in the Network Infrastructure forums, place hasn't been the same without you around.

Hope you are well and thanks for the LinkedIn recommendations :)

Jon

[Giuseppe Larosa]

Hello Jon,

yes I will try to be present in the forums this year.

>> place hasn't been the same without you around.

You are too kind. There is still you, Peter Paluch, Joseph Doherty and others

I'm well and

>> thanks for the LinkedIn recommendations :)

they are well deserved and verified but your activities in the forums.

Best Regards

Giuseppe

[foysol_bgd]

Thanks for your reply Giuseppe. 

I implemented the solution and tune the BGP to 5 and 15 and left bfd and spf timers that I mentioned above. 

But I have another problem now :) 

I am using Ethernet transport for this network and there is a problem with end to end tracking for OSPF over Ethernet. Even though there is a link failure in the middle, OSPF did not re-converge over to another link. In brief, my router interface was up and ospf did not failover to a different path. 

I understand BFD will not help here. I believe I need to configure OAM or CFM any suggestions?

[Giuseppe Larosa]

Hello Foysol_bgd,

>> I believe I need to configure OAM or CFM any suggestions?

What platforms are you using?

I have used successfully MPLS OAM to monitor an EoMPLS pseudowire in the past some years ago between two C7600 with RSP720.

I currently have problems with ethernet CFM (Connectivity Fault Management ) over a VPLS where PE nodes are ME-3600-X. We have just opened a Cisco TAC SR for this and I'm waiting for feedback from the TAC engineer.

>> I am using Ethernet transport for this network and there is a problem with end to end tracking for OSPF over Ethernet. Even though there is a link failure in the middle, OSPF did not re-converge over to another link,

This is not clear to me, I understood you have configured Ethernet transport over MPLS and that you likely have OSPF running between L2 CE nodes over it. I don't understand the part where you say "Even though there is a link failure in the middle, OSPF did not re-converge over to another link,". Why it should do, OSPF exchanges OSPF hellos over the ethernet over MPLS network but it unawrare of faults in the backbone networks.

Hope to help

Giuseppe

Edit:

you are using OSPF as the IGP in your backbone, do you refer to this OSPF instance? but in this case BFD should make the job.So for this reason I have supposed you have another OSPF process/domain running between L2 CE nodes attached to the Ethernet over MPLS service.

[foysol_bgd]

Hi Giuseppe Larosa, 

Thank you for your reply. I need to check more about the configuration I did and need to do some trouble shooting as well. In brief, 

There is a network outage in Ethernet cloud in the above picture and the interface in Device A is UP to Device B. The outage happened in the middle of the cloud and not to the end devices connecting Device A and Device B. My understanding was BFD should be able to detect this and OSPF neighbhorship will form via Device C. 

I was able to ping Device B from Device A and traffic was routed via Device C but OSPF did not form the neighbhorship via Device C. I have used Loopback interfaces for OSPF. 

I need to check this more and make sure there is no configuration mistakes. 

[Giuseppe Larosa]

Hello Foysol_bgd,

>> There is a network outage in Ethernet cloud in the above picture and the interface in Device A is UP to Device B. The outage happened in the middle of the cloud and not to the end devices connecting Device A and Device B. My understanding was BFD should be able to detect this and OSPF neighbhorship will form via Device C. 

For sure BFD will detect the failure and will make OSPF to re-run SPF and to use device C to reach device B.

>> I was able to ping Device B from Device A and traffic was routed via Device C but OSPF did not form the neighbhorship via Device C. I have used Loopback interfaces for OSPF. 

This is not clear to me, if you can ping the device B after the fault in the ethernet cloud, this means the OSPF adjacency between Device A and Device C was already in place or am I missing something?

>> I have used Loopback interfaces for OSPF. 

I hope you are not using unnumbered interfaces referring  to loopback to receive an IPv4 address.

You should use the loopback address just as OSPF routed.ids, each physical link has its own IPv4 subnet. You can use private IP addresses in RFC1918 for this you don't need to use public IPv4 addresses for infrastructure links.

Hope to help

Giuseppe

[foysol_bgd]

I am very sorry and I think I have over complicated. Appreciate your comments. 

Would like to know when the Device A to Device B link via the Ethernet cloud is down, how can I make my router interfaces down as well?

I know BFD will track the neighbhor but I like to make my interface status to down/down. What will be the best way?

[Giuseppe Larosa · Accepted solution]

Hello Foysol_bgd,

>> I know BFD will track the neighbhor but I like to make my interface status to down/down. What will be the best way?

An EEM script can do the job, post your request under the appropriate forum

(Network Infrastructure> EEM scripting)

where Joe Clarke and others can help.

However, from a routing point of view the physical state of interfaces to the ethernet cloud is meaningless, because BFD make the OSPF adjacencies to be turned down over the ethernet cloud and all routers recalculate the SPF to avoid the broken link.

So you don't really need the EEM script to make your network work as the OSPF + BFD takes care of any topological change including an issue inside the ethernet cloud.

So my suggestion is that you don't need to worry about the interface state, focus on the convergence time for the IP subnets of the services if they are quick as BFD with small losses you are fine.

Hope to help

Giuseppe