Cisco 350-401: BGP Path Selection in Practice
On this page
BGP path selection is a policy decision performed on a set of eligible routes, not a promise that the path with the lowest latency will win. A router may receive several paths to the same prefix through different neighbors; attributes, route policies and implementation rules determine which one becomes the selected route. Engineers can spend hours checking physical links when the actual cause of an unexpected next hop is a changed local preference, an inbound filter, or a prefix that was never eligible in the first place. Troubleshooting should therefore reconstruct the decision from the candidate routes rather than memorize a generic best-path list.
Establish which routes are actually eligible
Start with one destination prefix, the receiving router, and its BGP address family. Determine which neighbors are established and whether the prefix was advertised, accepted, or filtered. A route that does not appear in the relevant table cannot win best-path selection. Similarly, a path whose next hop is not reachable may be excluded or fail to install, depending on platform and configuration.
Distinguish three observations: what a neighbor advertised, what the local BGP process retained as candidates, and what the routing and forwarding tables ultimately installed. Those are not always identical. Route reflection, policy modification and next-hop resolution can change the visible result at each boundary.
Read attributes in policy context
Cisco BGP implementations use an ordered selection process whose details vary by platform and release. Attributes such as local preference, AS path, origin, MED, and the type of BGP peer can influence the choice after earlier eligibility checks. Cisco-specific weight is local to a router and should not be treated as an attribute propagated across the whole autonomous system.
Local preference is a common way to express a network-wide preferred exit within an autonomous system. AS-path prepending may influence inbound selection by external neighbors but is not a guarantee, because their policies can override it. MED is commonly relevant when comparing routes from a neighboring AS under applicable conditions, not a universal substitute for local preference. Confirm the actual platform’s decision order and any configuration changing comparison behavior before attributing a result to one field.
Work through a reproducible example
Imagine a branch has two external connections, Provider A and Provider B. Both announce reachability to an application prefix. The network team intends to send traffic through A, but B is selected after a policy change. Capture both received candidates, including the source neighbor, local preference, AS path, next hop and whether the path is considered valid and best.
If A’s path is absent, investigate advertisement, inbound prefix lists and route policies before tuning attributes. If both paths are present, identify the first decision criterion that distinguishes them. If the selected BGP path is correct but packets still use a different route, inspect administrative distance, route recursion, the actual route installed in the RIB, and applicable forwarding or policy-based-routing behavior. A clean route-policy diagnosis should explain the observed forwarding path, not just the BGP table.
Change only the policy that owns the decision
Make a bounded change in an authorized lab: adjust a route-map value on one test prefix or restore a mistakenly applied route filter. Record the expected winner before committing the configuration. Avoid modifying several attributes simultaneously, which makes the reason for the new outcome unclear and increases the risk of a surprise return path.
Verify the result at both ends where relevant. A preferred outbound path does not automatically control return traffic selected by remote networks. Check whether the application requires symmetric treatment, whether stateful inspection sits on a particular route, and how the solution behaves if the preferred provider becomes unavailable.
Turn the best-path algorithm into operational evidence
For an engineer practicing Cisco 350-401 ENCOR routing scenarios, reading BGP attributes is useful only when those facts support a diagnosis: which routes were available, which policy affected them, why a particular route won, and what the data plane did as a result. Document this reasoning rather than presenting a memorized ordered list without the specific platform’s conditions.
A solid troubleshooting note identifies one prefix and VRF, the relevant neighbors, eligible candidates, the first decisive difference, the installed route, and the evidence that the intended traffic follows it. That approach scales from a lab to a multi-site production network without confusing configuration intent with measured behavior.