Forum Discussion
NimbostratusOSPF Route Advertisement for Floating Self IP in Active/Standby HA
MVPWhat is the exact issue you are having?
I have similar setup for BGP F5 advertising the floating ip is much better than devices where the standby is "death" and only when the active goes down then OSPF or BGP sessions need to be established.
The new active F5 will send gARP for the floating IP. Some network environments dislike or drop gARP and then Configuring MAC masquerade (11.x - 21.x) is used as then the F5 devices use vMAC that will failover to the new active F5.
You need to properly test and know your network environment.
LBJThank you for your explanation!
I have one more question.
If the Floating IP is advertised, will OSPF/BGP route convergence still affect traffic during an HA failover?
My understanding is that after failover, the new active F5 sends a Gratuitous ARP, allowing the upstream switch to update its MAC table and forward traffic to the new active unit.
Does this mean traffic can continue without waiting for OSPF/BGP convergence?
- Nikoolayy1
Yes that is the idea that there is no convergence on routing protocol side as only layer 2 with gARP or Mac masquerade needs to happen.
You need to get used to the idea that the next hop in the routing advertisements is the Floating IP but the local self-ip are used for bgp or ospf neighbors.
For bgp it is different as only the active system advertises the floating-ip as next hop for VIP addresses but OSPF has a need to convergence Support for the OSPF routing protocol in BIG-IP LTM HA pairs .I suspect that even without floating IP things will work as then the standby will advertise with different cost the local self-ip.
See Lab 1: West Data Center Configuration (there no the same article for OSPF 😥 ) as I tested and saw that for myself Floating-IP with Mac masquerade as the best option but that depends of the networking environment. Interesting with BGP even if the active system only advertises the Virtual server VIP addresses, if you configure default route to be advertised from F5 to the downstream routers then the Standby will advertise the default route as well but with next hop the floating-ip, so again traffic will be send to the Active system not standby.
Other optimization that you could test Overview Of BFD Protocol but the switch/route people will know the best if they want this as well.
For me F5 is made to be integrated with ospf or bgp as nowadays for example Kubernetes k8s clusters use this routing protocols in many cases but just never select F5 to be the neighbor of 100 servers as the fabric between F5 and kubernetes cluster should be Layer 3 not Layer 2 and to be routing neighbor of the F5 and servers (the next hop if you will). This way F5 will have 1 or 2 neighbors (not 100 that F5 maybe is not made for) and server changes will trigger routing config update on the fabric side not F5. No problem for F5 pool members to be 1 hop away and not directly connected on layer 2.
