Routing

 View Only
last person joined: 3 days ago 

Ask questions and share experiences about ACX Series, CTP Series, MX Series, PTX Series, SSR Series, JRR Series, and all things routing, including portfolios and protocols.
  • 1.  BFD and IGP

     
    Posted 04-15-2021 07:37
    I just need some clarity regarding the role that BFD plays along side IGP for example ISIS or OSPF. Does BFD still work independently and only signals to the protocol that it detected a failure and therefore should also check if they (the protocol) experience the same. For example BDF recieves no response, tells OSPF please ignore your timers and send a keep alive now. Should the protocol receive nothing then it goes into the down state should it still revieve is response from it neighbor it will remain up although BFD is down.


  • 2.  RE: BFD and IGP

    Posted 04-16-2021 07:18
    I'm not sure I follow the question, so sorry if this is not what your looking for.

    BFD runs separately and when it detects an issue based solely on BFD timers and status it tells the partner routing protocol to immediately tear down and withdraw routes.

    Then the partner protocol starts its normal process to connect with neighbors which will work from step 1 of start up process and continue until the connection is restored.

    ------------------------------
    Steve Puluka BSEET - Juniper Ambassador
    IP Architect - DQE Communications Pittsburgh, PA (Metro Ethernet & ISP)
    http://puluka.com/home
    ------------------------------



  • 3.  RE: BFD and IGP

     
    Posted 04-16-2021 08:44
    We had a strange issue where a link over a 3rd party service would  black hole IP traffic yet the IS-IS would remain up. So i did a virtual LAB setup where i would setup the same and also enable BFD along side the IGP. I would have a transparent switch which the link between the two routers are connected and i created a firewall filter to simulate dropping certain traffic types. So running both BFD and IS-IS i started dropping unicast traffic on the link  only, this brought the BFD down but the IS-IS remained up, doing wireshark i could still see the LSA  going through on multicast between IS-IS neigbours and the timers where still been updated. So even when BFD failed the protocol remained up. i reset the test and then did the reverse, the protocol went down but BFD remained up, this seems normal. Did another test where i applied the filter to drop all traffic right at the start of a new IS-IS keep alive and then it would go down immediate. So this tells me that BFD actually only signals the IGP that there is possible failure and therefore need to immediately check the same using its own mechanisms for failure detection, if its own mechanism detect no failure then it will continue operating normally.


  • 4.  RE: BFD and IGP

    Posted 04-16-2021 13:00
    I saw similar in my lab (albeit Cisco XRv9K) when running BFD for OSPF.... (I'll try to test Juniper vMX later)

    First, sniffer shows OSPF Hellos (mcast) running alongside BFD (ucast) (Note 1)

    Test 1 - I disabled BFD on one side, and OSPF Neighbor stays up, BFD of course went down on the far side immediately

    Test 2 - I have a lab (virtual) that doesn't have physical links between routers, as they are software-based connections, point is, link/carrier stays up on far side even though I shutdown one side of the link... point is, I can simulate as if i had a link outage on one side, and the other side thinks it's still up... so with that said...

    I shutdown one side...

    ...the far side bfd immediately goes down, and, ospf immediately goes down.  Good.  I like it.  BFD seems to communicate to OSPF immediately that it needs to go down.

    Test 3 - I removed BFD completely from OSPF config (both sides).  same test as Test 2.  Shutdown one side of link, and, as you might expect, the far side OSPF took until the deadtime (~40 seconds) to finally drop neighborship.

    So, BFD, as advertised, seems to work nicely in this case (particularly Test 2, when enabled with a client protocol like OSPF)

    - Aaron

    Note 1 - BFD ucast is quite interesting, as RFC 5881 states that the ip address in the header should be such that the receiving system routes the traffic back to the sender, with that said, it's the same ip address for src and dst ip! I thought that was strange until i saw it in the rfc (see section 4. Encapsulation, 2nd paragraph).  of course the underlying mac address flips to reflect correctly, but the src and dst ip's do not change

    https://tools.ietf.org/html/rfc5881

    i also learned of s-bfd (seamless bfd) for use with segment routing

    ------------------------------
    Aaron Gould
    Senior Network Engineer
    aaron@gvtc.com
    https://www.linkedin.com/in/agould123/
    ------------------------------