R80.20 Gaia ClusterXL HA BGP Routing Causes Outage During Failover

[Dave_Cullen]

Hi All

I have configured BGP between an R80.20 ClusterXL HA cluster and a single SD WAN “Socket” device from CATO Networks.

I have neighboured the Socket to the ClusterXL VIP as described in one of the Check Point SKs I have been following.

I do see all learned routes on both gateways when issuing show route in CLISH.

howwver when I fail over it causes a drop of 3 to 6 pings.

It’s not massive I know, but I am used to ClusterXL failover being seemless up until now.

SK100499 (primarily a graceful restart SK but it shows my similar BGP config in the screenshots)

Ive looked in to BGP graceful restart but this isn’t supported by the CATO Socket right now.

Short of adding static routes instead... does anyone know what else I can do to stop that delay for routes to re-learn and pass traffic again?

Many Thanks

Dave

[PhoneBoy]

My understanding is that routing should be synced between the cluster members provided the router ID is the cluster IP address on both members. What are you using in this case?

[Dave_Cullen]

Hi PhoneBoy

Yes indeed the outer ID is the cluster VIP of the interface where the peering takes place.

i am seeing the routing sync, but it’s that outage during cluster failover that I have an issue with.

Apparently graceful restart is in the CATO roadmap, but I wondered if there was a workaround I could apply on the checkpoint to help in the meantime. 

Cheers