When configuring network redundancy or advertising new routes from On-Prem to VMware Cloud on AWS SDDC, learned routes for specific subnets (often larger CIDR blocks like /24 or /16) fail to be programmed into the NSX Edge Tier-0 routing table.
The following symptoms are observed during this issue:
VMware Cloud on AWS
This issue is caused by a routing conflict triggered when identical prefixes are advertised simultaneously from both Transit Gateway (TGW) and Direct Connect (DX).
The routing engine encounters an inconsistency during route programming, causing the NSX Edge T0 router to reject the update for that specific prefix.
To resolve this conflict, you must ensure that routing advertisements are unique and non-overlapping across the TGW and DX paths.
1. Identify Duplicate Routes: Review your BGP route advertisements on both the on-premises Router and the AWS TGW.
2. Remove Overlaps: If a prefix is being advertised from both TGW and DX simultaneously, remove the advertisement from one of the two paths so that the route is learned via a single preferred source only.
3. Once duplicates are removed, the NSX Manager should automatically retry the route programming.
4. Monitor the SDDC NSX Manager UI to confirm the route state changes to `SUCCEEDED`: Navigate to the NSX Manager UI -> Cloud Services -> Direct Connect or Transit Connect -> Learned Routes tab and verify that the Network is now present in the Learned Routes.
Note: If a redundant route is required via both TGW and DX, as a work-around you could advertise a more specific route via your preferred path (e.g. /25 via DX) and a less specific route via the backup path (e.g. /24 via TGW).
If you require urgent assistance with mitigation or additional clarifications on this issue, please raise a Technical case with Broadcom Support for the Product VMware Cloud on AWS (Get Support) with the following details:
VMC Engineering is working on a permanent fix for the issue in a future release. Currently there is no ETA for the fix availability.