Configure QoS for BFD packets

[maverick0]

Hey Guys,

I am here to request some help because I need to configure a QoS policy to avoid BFD session goes down. I know that BFD packets are marked as CS6 because I have been performed a packet capture to confirm that information. 

Basically, the BFD session goes down when the link are facing a high utilization. Currently, there is no QoS policies applied on the routers, but I need to avoid this impact.

If someone here already faced this issue, could share some information how to solve it?

Thanks in advance.

[balaji.bandi]

It depends on the Router you are using -

Most of the recent routers have control plans that have priority, how big is your link ? you can cap 95% bandwidth utilization and prioritise

check below thread suggestion :

https://community.cisco.com/t5/routing/prioritize-bfd-packets/td-p/2136263

[MHM Cisco World]

can you share the BFD config you use ??

[Joseph W. Doherty]

The "link" in question is a direct physical link between the routers, or do the two routers connect to a switch, which can also direct other port traffic, on that switch, to the router ports?  If the latter, you might also need to "do" QoS on the switch too.

You might first try, on the router ports, something like:

policy-map tryThis
class class-default
fair-queue

interface # !bfd interface
service-policy output tryThis

[maverick0]

The sites are connected using a DWDM infrastructure and they have a minimum of 2Mbps or 8Mbps depend of the site location. The router's interface is connected into the WAN device.

In the ASR903 platform, I have enabled the BFD in the OSPF process using this command:

router ospf 1

 bfd all-interfaces

and the BFD internal and multiplier are enable on the interface level, as following below:

interface GigabitEthernet0/3/4

 bfd interval 50 min_rx 50 multiplier 3

Here are a sample of the log entries for BFD session that start to happens when the link are in high utilization conditions.

*Sep 29 15:52:02.891 UTC: %BFD-6-BFD_IF_CONFIGURE: BFD-SYSLOG: bfd config apply, idb:GigabitEthernet0/3/4
*Sep 29 15:53:00.731 UTC: %BFD-6-BFD_SESS_CREATED: BFD-SYSLOG: bfd_session_created, neigh 172.16.1.180 proc:OSPF, idb:GigabitEthernet0/3/4 handle:1 act
*Sep 29 15:53:00.776 UTC: %BFDFSM-6-BFD_SESS_UP: BFD-SYSLOG: BFD session ld:1 handle:1 is going UP
*Sep 29 15:53:05.805 UTC: %BFD-6-BFD_SESS_CREATED: BFD-SYSLOG: bfd_session_created, neigh 192.168.1.186 proc:CEF, idb:GigabitEthernet0/5/3 handle:2 act
*Sep 29 15:53:11.606 UTC: %BFD-6-BFD_SESS_CREATED: BFD-SYSLOG: bfd_session_created, neigh 192.168.32.65 proc:CEF, idb:GigabitEthernet0/4/4 handle:3 act
*Sep 29 15:53:14.238 UTC: %BFDFSM-6-BFD_SESS_UP: BFD-SYSLOG: BFD session ld:3 handle:3 is going UP
*Sep 29 15:53:23.052 UTC: %BFD-6-BFD_SESS_CREATED: BFD-SYSLOG: bfd_session_created, neigh 192.168.3.82 proc:OSPF, idb:GigabitEthernet0/5/6 handle:4 act
*Sep 29 15:53:23.063 UTC: %BFDFSM-6-BFD_SESS_UP: BFD-SYSLOG: BFD session ld:4 handle:4 is going UP
Sep 29 16:05:11.633 UTC: %BFDFSM-6-BFD_SESS_DOWN: BFD-SYSLOG: BFD session ld:4 handle:4,is going Down Reason: DETECT TIMER EXPIRED
Sep 29 16:05:11.634 UTC: %BFD-6-BFD_SESS_DESTROYED: BFD-SYSLOG: bfd_session_destroyed, ld:4 neigh proc:OSPF, handle:4 act

[Joseph W. Doherty]

Ok, unless you're sure high utilization is only in one direction, you may need to insure all interfaces, from BFD interface to BFD interface have QoS.  I.e. QoS might be needed on your ASR903 and/or you noted "WAN" router it connects to.

[maverick0]

@Joseph W. Doherty 

Thank you for your reply.

The DWDM is transparent for the routers as if they are directly connected.

As the BFD packets are self-generated by the local router and sent over the link. My question is how to prioritize them when they reach the other router. There is a specific command in the policy-map as you have suggested?

[Joseph W. Doherty]

What does g0/3/4 for neighbor(s)?  I.e. p2p or multi-point?

[maverick0]

Yes, OSPF point to point.

interface GigabitEthernet0/3/4
bandwidth 8192
ip pim sparse-mode
ip ospf network point-to-point
ip ospf hello-interval 15
ip ospf 1 area 0
negotiation auto
bfd interval 50 min_rx 50 multiplier 3
end

[Joseph W. Doherty]

Sorry, should have picked up on this earlier.  If you running across a WAN, which site-to-site links offer 2 to 8 Mbps, but your physical interface can transmit faster, it's very, very easy to have congesting in the SP's WAN cloud and drop packets there.  I.e. for QoS to be effective, you would need to shape to WAN's bandwidth.

At the moment, it's unclear to me what your logical and physical topologies look like.

What do you have, one gig interface (same device providing g0/3/4?) per remote site?  (Each a different DWDM channel?)

[MHM Cisco World]

this from CCIE slides, which recommend setting of BFD, you use 50ms and I think that is problem. 

the issue I think not link utilize but CPU time to analysis the BFD.

ONE more link about BFD and recommend timer.

http://brbccie.blogspot.com/2014/06/everything-bfd.html

[maverick0]

@Joseph W. Doherty  and @MHM Cisco World 

Thank you for your replies.

Let me try to change the BFD interval based on this suggestion and see what happens.