JNCIS-DC common mistakes
JNCIS-DC · Understand the traps and how to avoid them.
Memorizing EVPN route type numbers (Type 2, Type 3, Type 5) without being able to say which MAC/IP, IMET, and IP-prefix information each carries, so questions that describe a scenario (e.g., 'which route type advertises the anycast gateway MAC') cannot be answered by recall alone.
Candidates build flashcard sets around route type numbers because they are easy to quiz, then never connect them to the Layer 2/Layer 3 forwarding behaviour the exam actually tests. The numbers are the shallowest layer of the topic.
What to do instead
For each EVPN route type, write one sentence describing the NLRI key and one describing what a PE does on receipt. Then trace a host-to-host flow across two leafs and label which route type supplies each lookup. If you cannot place the route type in the flow, you do not know it yet.
Treating VXLAN VNI and VLAN ID as interchangeable, or assuming a 1:1 mapping is mandatory, when the design uses a VLAN-to-VNI mapping table with many-to-one or one-to-many relationships.
Lab setups often use a single VLAN and single VNI, which hides the mapping layer. Candidates then answer questions about multi-tenant designs as if the two identifiers were the same field.
What to do instead
Draw the encapsulation stack (outer IP, UDP, VXLAN header with 24-bit VNI, inner frame with 802.1Q tag) and mark where each identifier lives. Practise a scenario with two VLANs mapped to one VNI and one VLAN mapped to two VNIs, and state what the egress VTEP does in each case.
Configuring an EVPN-VXLAN fabric but omitting or misplacing the underlay routing and loopback reachability, then being unable to explain why BGP EVPN peering stays down even though the overlay configuration looks correct.
Candidates jump straight to the overlay because that is where the interesting configuration lives, and forget that EVPN peering rides on loopback-to-loopback reachability that the underlay must provide.
What to do instead
Adopt a fixed build order: underlay IP addressing, loopback interfaces, underlay routing (e.g., OSPF or EBGP), verify loopback-to-loopback ping, then overlay peering. When troubleshooting, always confirm the underlay first and write down the ping test you would run.
Answering questions about MC-LAG and ESI multihoming by listing redundancy benefits without distinguishing how the two mechanisms handle split-brain, per-flow hashing, and failure of the inter-chassis link.
Both provide active-active server attachment, so candidates collapse them into one concept. The exam probes the failure behaviour, where they differ substantially.
What to do instead
Build a comparison table with rows for control protocol, data-plane forwarding on link failure, split-brain detection, and orphan-port behaviour. For each row, describe what happens when the peer link fails, because that is the scenario most likely to be tested.
Confusing a vPC or MC-LAG peer link with a regular trunk and assuming it carries ordinary server traffic, leading to wrong answers about which path traffic takes and what happens when the peer link saturates.
The peer link is cabled like a data link and carries some data in failure scenarios, so its dual role is easy to misread. Candidates also miss that orphan ports and peer-link failure have defined behaviours.
What to do instead
State explicitly, in writing, the primary role of the peer link (control and keepalive plus backup data path), then describe the three states: normal, peer link down, and peer switch down. Note which ports are suspended in each state.
Memorizing spanning tree timer values and port-state names without being able to reason about a topology change: which bridge becomes root, which port is alternate versus backup, and how a topology change notification propagates.
Timers are easy to memorize and appear in questions, so study time goes there. Topology-change reasoning requires drawing, which candidates skip.
What to do instead
Take a five-switch triangle-and-square topology and derive root election, root port, designated port, and blocking port by hand, showing the cost arithmetic. Repeat after failing the current root. Do this until the derivation is mechanical.
Reading a routing table or BGP output in a question and answering from the first matching line rather than applying longest-prefix match, administrative distance, and then metric in the correct order.
Candidates learn the tie-break rules as a list but do not rehearse applying them in sequence under time pressure, so they stop at the first plausible route.
What to do instead
For every routing-table question, write the three-step ladder explicitly: longest prefix wins outright; among equal prefixes, lowest administrative distance; among equal distance, lowest metric. Then apply it line by line to the presented table before choosing.
Assuming a firewall or security policy question is about the policy match criteria when it is actually about session state, zone assignment, or the order in which policies are evaluated, and answering with the wrong layer of the device.
Policy configuration syntax is the visible part of the topic, so candidates study rule syntax and overlook stateful inspection behaviour and zone binding, which determine whether traffic is even evaluated.
What to do instead
For each security scenario, first identify the ingress and egress zones and confirm the interface is bound to the right zone, then check session state, then the policy. Practise explaining why a permitted policy can still drop traffic because of asymmetric routing or a missing return session.
Treating automation and telemetry topics as vocabulary recall (NETCONF, RESTCONF, gNMI, YANG) without knowing which transport, encoding, and data model each uses, so scenario questions about choosing a mechanism are answered by keyword association.
These are acronym-dense topics and candidates memorize expansions rather than the architectural differences, which is exactly what scenario questions test.
What to do instead
Build a four-column table: protocol, transport, encoding, and data model, and add a column for typical use (configuration versus streaming telemetry). Then answer three 'which would you choose for X' scenarios using the table as justification.
Running out of time because early questions are re-read repeatedly, leaving multi-part scenario questions at the end unanswered or guessed.
Scenario questions are long and candidates try to fully solve each one before moving on, which is a poor strategy when the exam mixes short recall with long analysis.
What to do instead
Do a first pass answering only questions you can resolve in under a minute, flag the rest, then return. Before the exam, rehearse with a timed set that includes long scenario items so the pacing decision is practised rather than improvised.