load balancers convergence (Radware)

Mihai mihaigabriel at gmail.com
Thu Dec 8 22:20:35 UTC 2016


  Sorry if this is not the right list to post but it's the last resort 
and any clue would be highly appreciated.
I am new to LBs world and have the following active-standby topology 
(Alteon 5524):

|                       |
|                       |
eBGP                    eBGP
|                       |
|                       |
Alteon1(act)    Alteon2(stb)
|                       |
     WEB servers

- BGP local-pref is higher on the R1-A1 session.
- VRRP priority is higher on Alteon1.
- both Lbs advertise the same VIP address in BGP.
- NAT is configured for WEB servers IPs.

I am using these devices to load balance HTTP traffic and have the 
following issue:

1. Alteon1 fails and the traffic moves to Alteon2 without problems.
2. After Alteon1 recovers, it becomes the VRRP master due to higher 
priority, but the BGP session between Alteon1 and Router1 establishes 
after VRRP preemption (more than 1 minute after A1 becomes master) and 
the traffic gets dropped. I tried to use the hold-off timer to delay the 
VRRP preemption to match the BGP session establishment but still have 
~30s downtime.
Creating a direct link and BGP session between Alteons does not help as 
the traffic will be asymmetrical and is dropped on Alteon1.


More information about the NANOG mailing list