Forum Discussion

Thiyagu's avatar
Thiyagu
Icon for Cirrus rankCirrus
Jun 17, 2020

ARP issues between Nexus 7K and F5 LTM

Hi Team,

We have faced a weird issue in our environment.

issue:

loadbalancer-1 failedover to Load balancer-2. Whereas all the production traffic flow did not failover properly to load balancer-2 for around 25 minutes. After 25 minutes load balancer-2 started getting production traffic to the VIPs.

 

Does anyone have experienced similar ARP issue between F5 LTM and Nexus 7K at the time of fail-over?

Is there any workaround to this issue meaning any configuration change on LTM and the time of issue?

Is there any configuration change on LTM which would help to trigger GARP request two or three time to the upstream N7K?

 

Regards,

Thiyagu

application delivery
BIG-IP
LTM

2 Replies

Sort By
  • Henric_Petterss's avatar
    Henric_Petterss
    Icon for Cirrus rankCirrus
    Jun 17, 2020

    Don't know if it is the problem here, but F5 LTM will not send out GARP for vlans where it doesn't have a self-ip. We noticed that recently when we got the same problem as you have. We did not have self-ip in all vlans so the VIP's on that were not moved in a failover. We added a self-ip on that vlan (in the matching ip-subnet), and then failover started working again.

     

  • Skyler's avatar
    Skyler
    Icon for Nimbostratus rankNimbostratus
    Jul 24, 2024

    Hi Thiyagu,

    Just to confirm: how many IPs do you use to send GARP?


    Our team has currently found an issue similar to yours, during failover, production traffic is not moving correctly to the primary device, through packet capturing we found out that it is the F5 sending huge amount of GARP packets during failover, which triggers the CoPP of the N9K, which drops the excessive GARP packets, resulting in the MAC table not being updated, and all the traffic returns are still being sent towards the standby device.
    In 25 minutes time, that's when the MAC table ages before N9K correctly learns the correct new MAC.

    A few of our practices include:
    Modify the value of CoPP on the N9K, the default value is 1400kbps.
    Limit the number of GARPs sent per second by F5, I think articles K11985 and K7332 can help.
    Please note that these are just small practices that apply to me, please carefully consider and think about the feasibility of the specific operation.
    If you have other things to share you can let me know and we can discuss it together.🙂