BGP Next-Hop Resolution Through Another BGP Route Can Mark Routes Inaccessible
Session Smart Router (SSR) may mark a Border Gateway Protocol (BGP) path inaccessible when its next hop resolves through another BGP-learned route, and select a different path.
Advisory Type: Field Notice
Issue ID: I95-66848
Last Updated: 2026-10-28
Introduced in SSR Version: 6.2.11, 7.0.5, 7.1.6, 7.2.0, 7.2.2, 7.2.3
Resolution Planned for SSR Version: Tracked by I95-66848
Problem
SSR's BGP next-hop tracking now rejects paths whose next hop resolves through another BGP-learned route. This prevents recursive routing feedback loops, but can also invalidate legitimate paths in designs that rely on BGP-learned next-hop reachability.
Symptom
An affected path may appear in BGP output as inaccessible.
When another valid path exists, SSR may select that alternate path, potentially changing traffic direction, path preference, or failover behavior. If no eligible alternate exists, the destination may have no usable route. The impact depends on the network topology and available paths; not every BGP route or deployment is affected.
Representative output:
show bgp <prefix>
<prefix> <next-hop> (inaccessible, import-check enabled) from <BGP-peer>
Origin <origin>, localpref <value>, invalid, internal
Cause
The next-hop tracking check was introduced to prevent recursive route-resolution feedback between BGP and zebra, SSR's routing-table manager. In a feedback loop, a BGP route becomes unreachable, an alternate route changes the routing table, and the original next hop then appears reachable through a BGP route whose installation depends on that same routing state. BGP and zebra can repeatedly update one another, delaying convergence and route withdrawal during failures.
To prevent this condition, SSR rejects BGP routes whose next hops resolve through BGP-learned routes. That guard also rejects some valid designs where a distinct BGP-learned route provides reachability to a route's next hop.
Workaround
On the SSR where the affected destination route is received, run show bgp <destination-prefix> to inspect its paths. Identify the affected path's NEXT_HOP, then check how SSR reaches the next-hop address.
This can happen when a route's NEXT_HOP is a third-party BGP peer, and another SSR advertises the BGP route used to reach that peer.
If traffic can safely use the advertising SSR as an alternate transit path: Configure next-hop-self on that SSR's BGP session toward the SSR receiving the affected destination route. This replaces the original next hop with the advertising SSR's peer address, causing affected traffic to transit that SSR. Use this workaround only if the receiving SSR can reach the advertising SSR's peer address directly and the advertising SSR has a route to the destination. Do not use it if traffic must bypass the advertising SSR.
Generic SSR configuration pattern:
config authority router <ADVERTISING_ROUTER> routing default-instance routing-protocol bgp neighbor <RECEIVING_ROUTER_PEER_ADDRESS> address-family ipv4-unicast next-hop-self true
Apply the setting on the advertising router, not on the receiving router's outbound session. Verify actual peer addresses, address family, and outbound policy before changing configuration.
If traffic must continue to use the original next hop: Provide a valid route to that next-hop address through a non-BGP source, such as a connected or static route, where supported and appropriate for your design. Ensure this route is independently reachable, does not depend on the affected BGP path, and has correct failover behavior. Avoid creating a static or IGP recursion loop or unintentionally overriding a preferred route.
After applying the workaround, verify that:
- The receiving router sees a valid next hop resolved through the intended route source.
- The path is valid and selected as expected.
- The destination route is installed in the routing information base (RIB) and forwarding information base (FIB) through the intended peer.
- Bidirectional application traffic follows the intended path, including existing sessions and relevant failover scenarios.
Resolution
This behavior change is being addressed under issue I95-66848. Contact Juniper Support for release availability.