The route exists.
Does the service work?

Where do routing table entries come from?

A router may learn networks from connected Layer 3 interfaces, configured static routes, or a dynamic routing protocol such as OSPF. These sources contribute candidate routes. The router evaluates eligibility, next-hop resolution, administrative preference and protocol-specific metrics before installing usable routes into the Routing Information Base (RIB). The forwarding plane uses installed forwarding information to actually move traffic.

Different operating systems and vendors implement their routing databases differently. The examples in this phase use familiar Cisco IOS/IOS XE conventions. A route entry is evidence of the planned next hop, not proof that the next hop, reverse path or destination service actually works.

Longest-prefix matching is not administrative distance.

Two separate decisions are often confused. During route installation, an OSPF and a static route advertising the same destination prefix compete as candidates. In a conventional Cisco configuration, the lower administrative distance (AD) wins when the route is usable. For example, static AD 1 generally beats OSPF AD 110 for an identical 10.70.20.0/24.

During packet forwarding, the router selects the longest matching prefix from the installed routes. A valid OSPF 10.70.20.128/25 beats the installed static 10.70.20.0/24 for destination 10.70.20.180, because /25 is more specific. At this stage AD doesn't override prefix length.

For additional vendor-specific detail, see Cisco's Administrative Distance explanation ↗, which distinguishes RIB selection from forwarding-table lookup.

Every route has prerequisites.

Connected networks ordinarily appear when an appropriate Layer 3 interface is operational. A static route names a destination prefix and the next hop or exit interface. A default route (0.0.0.0/0) is a fallback for destinations with no more-specific match; it does not override an available /16 or /24 just because it has a lower administrative distance.

When the selected next hop is an IPv4 neighbor, the router normally requires a valid Layer 2 neighbor mapping. A configured route whose next hop cannot be resolved or whose interface is unavailable may fail to install, or may be unusable according to platform and routing behavior. Always correlate route status, adjacency, interface and the forwarding path.

Inter-VLAN routing still needs a Layer 2 foundation.

Finance VLAN 10 (10.50.10.0/24) and Servers VLAN 20 (10.50.20.0/24) are separate broadcast domains. A multilayer switch can route between their VLAN interfaces (SVIs), or an external router can perform that role over an appropriate 802.1Q design. Each client uses an authorized gateway address and the correct local route.

Even if the route table correctly lists both connected networks, an access VLAN mismatch, an SVI that is down, an ACL, a firewall policy or a server with the wrong return gateway can still prevent communication. The job is to determine which part of the path actually fails.

What OSPF neighbors and LSAs actually tell us.

Open Shortest Path First is a link-state interior gateway protocol. Within an OSPF area, participating routers exchange link-state advertisements (LSAs), build a link-state database (LSDB) and run a shortest-path-first calculation to derive routes. A metric such as OSPF cost influences the selected path among eligible OSPF alternatives.

An OSPF neighbor being FULL is evidence of control-plane adjacency, not proof that 10.70.50.0/24 was advertised, selected into the RIB, forwarded correctly, or permitted through security policy. Verify the LSDB, RIB, next-hop resolution, return path and service separately.

NAT/PAT changes addressing, not the routing requirements.

In a typical outbound IPv4 NAT/PAT deployment, many inside clients share an approved outside address. The gateway translates the private source IP and often the source port for each session. The external destination IP and destination port remain unchanged in this simple example. More complex NAT and proxies can behave differently, so document the actual design.

These addresses are examples, not public endpoints to test. The translation table can indicate that a mapping was created, but it does not prove that a reply came back, passed stateful policy, reached the inside client or completed a TLS/application workflow.

A successful investigation checks the inside/outside interface roles, egress route, translated addresses, any firewall ACL/state, reverse route and the actual application result. Don't indiscriminately clear translation tables or open wide permit rules as a substitute for evidence.

DHCP relay bridges the broadcast boundary.

A new DHCPv4 client may broadcast DHCPDISCOVER because it doesn't yet have an IPv4 address. Such broadcasts don't naturally traverse a routed VLAN boundary. A configured DHCP relay on the client-facing Layer 3 interface can forward messages toward an off-subnet DHCP server.

In conventional DHCPv4 relay behavior, the giaddr relay-agent field helps the server identify the client network and direct replies to the relay. Scope selection can also depend on server policy or relay options, so giaddr is not an absolute guarantee of a particular scope. A relay target being configured says nothing about whether the scope has free addresses or whether clients completed DORA.

For standards background, see RFC 2131 — DHCP ↗. In our example, a correct result means the client receives an approved lease, gateway and DNS information, and actually uses the intended service—not merely that DHCPOFFER appeared.

Make the forwarding decision yourself.

Try each destination in the route-selection workbench. Notice when connected routes win, why a static route is installed instead of an OSPF route for the same prefix, and how an installed, more-specific OSPF /25 still wins in forwarding over an installed static /24. Then withdraw the /25 and observe the new best match.

It's not always the routing protocol.

Work through four synthetic operational tickets: OSPF route selection, NAT/PAT, DHCP relay and an asymmetric return path. Each contains three diagnostic choices. Every response includes its reasoning; you can revise answers and repeat the incident without saving anything.

show ip ospf neighbor → 10.10.10.2 FULL/- on Gi0/0; show ip route 10.70.50.0 → route not in RIB.

Gather enough evidence to justify the next action.

Useful read-only commands should answer a specific question. On a Cisco-style router, the following checks support diagnosis. Syntax, output, privileges and defaults vary by platform/software, and some commands may expose sensitive network information. Use only approved devices and authorizations.