SD WAN : BFD + OSPF

[RS19]

In Cisco SD-WAN want to understand how BFD & OSPF co-work together.

I understand that BFD is used to monitor the underlay. If it find that the remote end point is not responding it brings down the bfd session. Want to understand how this works along with the OSPF. OSPF as routing protocol has it own timers to bring up and bring down the OSPF neighbourship. Wanted to understand how this both works together ?

[MHM Cisco World]

Did you check 

Show bfd session 

MHM

[RS19]

Sorry missed it.
Checked it. it is 1000 ms which is 1 seconds.
Multiplier is 7.

So my understanding is that BFD waited for 7 seconds, and show down alert. Since it is polling for every seconds, after 2 seconds it detected it was up and then shown the up alert. 

Is my understanding right. 

[MHM Cisco World]

https://www.networkacademy.io/ccie-enterprise/sdwan/application-aware-routing-app-route-policy

Check this link to understand poll interval

MHM

[RS19]

Thanks.

1 question, Regarding the below point from the link are they referring to  IPSec tunnel.

The Multiplier value determines how many consecutive BFD probes must be lost before a vEdge router declares a tunnel as down, as illustrated in figure 3 below.

[RS19]

In continuation to the above, from the above analysis it was down for 9 seconds, 
7 seconds for the bfd change + after 2 seconds it came up So in total 9 seconds.

Also I wouldl like to highlight that during this time, OSPF neigh was never down. How is that possible ?
My understanding is that if BFD is down, it will tell OSPF to reconverge or recalculate & why OSPF neig didnt go down ?

Also it was reported that the users face 1 minute of impact. But from the logs only 9 seconds was down.

What else could have caused this & what else I need to check.

[Kanan Huseynli]

This part is hard to diagnostics just with little information.

Basically when BFD goes down, then remote next-hop (NH) is considered invalid, so OMP routes with this NH will be invalid. And if these OMP routes were the best in local router's OMP table (where BFD was lost), then it impacts local recalculation of OMP and recalculation of redistributed protocol (OSPF).

OSPF neighborship is never gone down, because on service side you still have alive neighbor. Just your routes are changing and there is recalculation is OSPF domain. OSPF re-convergence does not mean that OSPF neighborship will be down.

[RS19]

Thanks. Its getting clearer now for me.
Will OSPF recalucation will cause any down time ? 

[Kanan Huseynli]

Of course, detecting change in LSDB (OSPF LSA DB), scheduling SFP algorithm, re-running it, re-writting to FIB newer routes etc. take time. Normally, all of these dont require 1minute which you faced, but 1 minute loss does not mean that you lost 1 minute only for OSPF re-convergence.

Keep in mind that, if change in routing protocol resulted change in actual forwarding, it mostly makes down time.

[MHM Cisco World]

please just confirm are there default route under the vpn 0?
I think you confuse about IGP/BGP run in transport and service 
SDWAN BFD not effect ospf in service it effect OMP.
the service IGP end in vedge and route is redistribute into OMP

MHM

[RS19]

there is not default route under VPN0

[MHM Cisco World]

what the IGP you use for transport not service ?
MHM

[RS19]

sorry I am not sure. How to check this ?

[MHM Cisco World]

can I see vedge config 
you mention ospf and I think ospf is use in service not in transport and hence it not relate to SDWAN BFD 

MHM

[MHM Cisco World]

correct 

[Kanan Huseynli]

Poll internal is used for Application Aware Routing (AAR). In short, these 120 sec (which is 2 minutes as you said) is the value of last valid time frame for AAR. It does not have relation to BFD down - up events. Only multiplier affects down - up event. And I see that you understood sending BFD every 1 sec and if 7x is lost then BFD tunnel is down.