PFR on iWAN

[Jatinder SIngh]

Hello,

i have a situation -- deployed iWAN with a hub and a transit hub. one of the site with WAN links - MPLS & Internet (both on different routers)

we are getting constant packet loss on MPLS link but still all my traffic does not shifts completely to the Internet link.

Can someone please suggest what are default values/timers for MC to take a decision for injecting a route change on the remote site border router. How does MC works in this case where there is packet loss or fluctuation on MPLS link (for eg.)

[Renan Abreu]

Hi Jatinder,

The branch sites are also testing the links through their BRs using Smart Probes.

The default monitor time is 30 seconds, it can be changed with the command "monitor-interval".

[Jatinder SIngh]

Thanks. my query is around PFR taking. role of shifting traffic in case of packet loss / link fluctuations
will it shift all traffic to second link/router
note - we see EIGRP flap also because of link issue

[Renan Abreu]

When you mentioned that the link fluctuates, what does it mean? Do you actually lose the link? Does the link actually go down or you just see packet drops from time to time? (it looks like packet drops as even EIGRP is being lost).

iWAN is all about monitoring and failing over. If the criteria for the link to failover to a secondary path is matched, you will perform it within the monitor-interval period. If you need a quicker failover than 30 seconds, you can adjust it manually.

[Jatinder SIngh]

hello,
thanks for the reply
yes we see packet drops on MPLS WAN link (EIGRP status also keeps changing)
what's suggested monitor-interval to overcome the link fluctuations ? plus this should not cause issues on traffic pattern if we change monitor- interval and then link fluctuation further cause traffic shift here and there between 2 links

[Jatinder SIngh]

plus one last question, we should change monitor-interval only on impacted Remote site - BR
correct ?

[Renan Abreu]

That's a tough question on how fast you should monitor it, I guess the answer would be as fast as you want it to failover. The most common I see is usually 5 seconds (This changes how quickly you send probes on links with actual data, so more things change than just the failover time).

This command is on the master hub, so, unfortunately, it would apply to all your branches. This is usually used on specific traffic classes (dcsp) that you want a fast failover, never seen it used to overcome a problematic link.  

[paul driver]

Hello

Do you have any polices defined for the MC to learn from?
Is your MC and Border rtrs on the same rtr or separate?

[Jatinder SIngh]

hello, I have not applied any specific policy.
MC and BR was configured by APIC EM for iWAN. in my scenario, I have 1 MC, 2 BRs in DC and 2 BRs in remote site
issue occurs when primary link at remote site fluctuates.
my expectation is PFR should sense that and divert all traffic to second link but it does not happens

[paul driver]

Hello

TBH I have not used PrF the last time i had exposure to this was OER, And then for the MC to be able to learn prefix learning you needed to tell it to learn and monitor via a delay or threshold value.