Hard CCNA Interview Questions

I check show interfaces on both ends for input errors, CRC, runts, giants, output drops and resets. Output drops suggest congestion, so I look at utilisation and QoS. CRC errors suggest a bad cable, dirty fiber or faulty SFP, so I check optical levels with show interfaces transceiver. I also confirm both sides agree on speed and duplex and that the MTU matches.

Suspect an MTU mismatch. Small pings fit in one frame, but full-size packets are dropped when one link or the new switch uses a smaller MTU, or when ICMP is blocked so Path MTU Discovery fails. Test with ping using a size of 1472 and the do-not-fragment bit, compare interface MTU on both ends, and check jumbo frame settings. Fix by matching MTU end to end or adjusting TCP MSS on tunnel interfaces.

Situation: our office had to move 300 users from old access switches to a new stack in one weekend. Task: I owned the plan. Action: I documented every port, pre-built configs, ran a pilot on one floor, wrote a rollback plan and kept a call bridge open with the desktop team. Result: we finished two hours early with only three port issues, fixed on Monday morning. I learned that a tested rollback plan gives the whole team confidence.

EUI-64 builds a 64-bit interface ID from a 48-bit MAC. Split the MAC in half, insert FFFE in the middle and flip the seventh bit of the first byte. For MAC 0050.56AB.CDEF the ID becomes 0250:56FF:FEAB:CDEF. Routers can use ipv6 address 2001:db8:1::/64 eui-64. Hosts often prefer random privacy addresses instead, to avoid exposing the MAC.

NDP uses ICMPv6 messages. A host sends a Neighbor Solicitation to the target's solicited-node multicast address (ff02::1:ffXX:XXXX) and gets a Neighbor Advertisement with the MAC. Router Solicitation and Router Advertisement messages provide prefix and gateway. Duplicate Address Detection uses NS to check that an address is unique. Using multicast instead of broadcast reduces load on other hosts.

Overlapping address space means routing alone cannot work. Short term, use NAT on the interconnect so each side sees the other as a different translated range, for example twice NAT or policy NAT. Long term, renumber one network into a free range, which is cleaner but needs planning for DHCP, DNS, firewall rules and static IPs. I would also document the IP address management plan to prevent repeats.

I would reserve a large block, for example 10.0.0.0/16, and give each branch a summarizable /20 such as 10.0.16.0/20, 10.0.32.0/20 and so on. Inside each, assign fixed ranges per role: users, voice, wireless, servers, management and WAN links. This lets each branch be summarized as one route, keeps firewall rules simple, and leaves room to grow. I keep it in an IPAM tool.

The MAC address table, stored in CAM, maps MAC addresses to ports and VLANs. When it fills, as in a MAC flooding attack, new addresses cannot be learned and the switch floods unknown unicast frames out of all ports in the VLAN, which lets an attacker sniff traffic. Port security limiting MACs per port prevents this.

An SVI comes up only if the VLAN exists in the VLAN database and at least one access or trunk port in that VLAN is up and forwarding, and the SVI is not shut down. So I check show vlan brief for the VLAN, check that at least one port in it is connected, check trunks allow it, and run no shutdown on the SVI. On Layer 3 switches, ip routing must be enabled for routing.

I would split users into VLANs by function or floor, each with a /24 or smaller, put voice, wireless, printers, servers and management in their own VLANs, and route between them on a Layer 3 core with SVIs. DHCP scopes and helper addresses would be added per VLAN. I would migrate in phases with a rollback plan and move VLAN 1 out of use for both users and the native VLAN.

A private VLAN splits one VLAN into a primary VLAN and secondary VLANs. Isolated ports can talk only to promiscuous ports such as the gateway, while community ports can talk to each other and to promiscuous ports. It is used in hosting, DMZs and hotel networks to isolate customers on the same subnet without creating many subnets. It requires VTP transparent mode or VTP v3.

Root Guard on a designated port puts it into root-inconsistent state if a superior BPDU arrives, protecting the planned root bridge, so it is used on distribution ports facing access switches. Loop Guard stops a non-designated port from becoming designated when BPDUs stop arriving, such as with a unidirectional fiber fault. UDLD complements Loop Guard by detecting one-way links.

Multiple Spanning Tree (802.1s) maps many VLANs to a small number of STP instances, so a switch does not run one instance per VLAN as PVST does. Switches in the same MST region must share the name, revision and VLAN-to-instance mapping. It suits large networks with hundreds of VLANs and multi-vendor environments, and supports load sharing by making different instances root on different switches.

Load balancing is per flow based on a hash. If the method is src-mac and most traffic comes from one router MAC, all flows hash to one link. I change port-channel load-balance to src-dst-ip or src-dst-mixed-ip-port so many flows spread across members. If the traffic is a single large flow, it can never exceed one member, so faster links would be needed.

Frequent TCNs flush MAC tables and cause flooding. I run show spanning-tree detail to find the port and switch originating topology changes, then follow it hop by hop. Common causes are a flapping link, a host port without PortFast that changes state each time a PC reboots, or a failing NIC. I fix the link and enable PortFast on edge ports so they do not generate TCNs.

UniDirectional Link Detection sends probes to confirm a neighbour receives traffic in both directions. On fiber, one strand can fail so a port receives but cannot send, which may make STP unblock a port and create a loop. UDLD normal mode logs and marks the port, while aggressive mode err-disables it. It is commonly enabled on fiber uplinks together with Loop Guard.

Situation: I approved a change that set a new root bridge priority, but the engineer applied it on the wrong switch and the campus reconverged during business hours. Task: restore service and find the gap. Action: we rolled back in ten minutes, informed users and ran a blameless review. Result: we added device name checks to the change template, peer review of exact commands and a maintenance window rule for any STP change. There was no repeat incident that year.

EXSTART or EXCHANGE stuck usually means an MTU mismatch, because database description packets larger than the neighbour's MTU are dropped. I compare MTU with show interfaces and set ip mtu to match on both sides. As a temporary workaround, ip ospf mtu-ignore can be used, but fixing the MTU is better. Duplicate router IDs can also cause odd adjacency behaviour, so I check those too.

Another source with a lower AD may be providing the same prefix, such as a static route, so the OSPF route is not installed. Or a more specific route wins by longest match. The next hop may also be unreachable, or a distribute-list may filter it. I check show ip route for the prefix, show ip ospf database, and compare administrative distances.

Run OSPF over MPLS for the primary path. For the backup, use a floating static default route via the broadband link with AD higher than OSPF, for example 250, or run an IPsec tunnel with OSPF at a higher cost. Track the primary with IP SLA if the provider does not drop the link on failure, so the backup activates even when the interface stays up.

Both paths must have equal total cost to be installed as ECMP. I check show ip route for two next hops. If only one shows, I compare interface costs, which differ if bandwidth or the reference bandwidth differs. I also check maximum-paths. Even with ECMP, load sharing is per flow using CEF, so a single large flow uses one link only.

This is asymmetric routing: the firewall sees only one direction of the flow and drops it. Make the path symmetric by using one active exit per prefix, for example HSRP or a higher OSPF default-route cost on the backup, and advertise inbound routes consistently with the same preference. If both links must be active, pair the firewalls with state sharing or use NAT per exit so return traffic comes back to the router that sent it.

CoS is a 3-bit priority field in the 802.1Q tag, values 0 to 7, used only on Layer 2 trunks. DSCP is a 6-bit field in the IP header, values 0 to 63, which survives across routed hops. Common values: EF (46) for voice, AF41 (34) for video, CS3 (24) for call signalling and default (0) for best effort. Mark close to the source and trust markings only at controlled boundaries.

Policing drops or re-marks traffic that exceeds a rate immediately, with no buffering, so it causes TCP retransmissions; it is used inbound or by ISPs. Shaping buffers excess traffic and sends it later, smoothing bursts to a configured rate at the cost of delay; it is used outbound, for example to match a 100 Mbps subscription on a 1 Gbps port.

Choppy voice under load means congestion on the WAN. I make sure phones mark RTP as EF and the access switches trust phone markings, then on the WAN router create a policy with a low latency priority queue for EF limited to the needed bandwidth, CBWFQ for other classes, and shaping to the real carrier rate so queuing happens on my router, not the provider's.

DAI checks ARP packets on untrusted ports against the DHCP snooping binding table and drops ARP replies whose IP-to-MAC mapping does not match. This stops ARP spoofing and man-in-the-middle attacks where an attacker claims to be the default gateway. Static hosts need ARP ACLs. It requires DHCP snooping enabled in the same VLAN and trusted uplink ports.

802.1X lets a switch or access point block a port until the user or device authenticates. The supplicant is the client, the authenticator is the switch or AP, and the authentication server is a RADIUS server. EAP messages are carried over the LAN as EAPoL and then inside RADIUS. On success the RADIUS server can assign a VLAN or ACL. MAB is a fallback for devices without supplicants.

IPsec uses IKE (v1 or v2) to authenticate peers and negotiate keys, ESP to encrypt and authenticate data, and optionally AH for authentication only, which breaks with NAT. Algorithms include AES for encryption, SHA-256 for integrity and Diffie-Hellman groups for key exchange. Tunnel mode encrypts the whole original packet and is used for site-to-site VPNs.

Disable Telnet with transport input ssh, enable SSH v2, and move admin login to AAA with TACACS+ so each engineer has a personal account with command authorisation and accounting logs. Keep one local fallback account in a password vault. Restrict VTY access with an ACL to the management subnet. Roll out with automation, test on a pilot, then document it for the audit.

Move management to a dedicated VLAN or out-of-band network, ideally in a management VRF. Apply an access-class ACL on VTY lines allowing only the jump host subnet, disable Telnet and HTTP, keep SSH version 2 and HTTPS only if needed, and use AAA with TACACS+ for per-user login and logging. Add control plane policing, block management subnets at the user VLAN gateways and rescan to confirm the fix.

Physical ports include service ports for out-of-band management, distribution system ports that connect to the switch as trunks, and redundancy ports for HA. Logical interfaces include the management interface used for CAPWAP and admin access, dynamic interfaces that map WLANs to VLANs, and a virtual interface used for web auth and DHCP relay. A LAG bundles distribution ports.

The underlay is the physical IP network of switches and routers that provides basic reachability, for example OSPF or IS-IS on routed links. The overlay is a virtual network built on top using tunnels such as VXLAN, so logical segments can stretch across the fabric independent of physical topology. SD-Access and SD-WAN use this model, with a fabric controlling the overlay.

I check the AP gets a DHCP address and can reach the WLC management IP. It discovers the controller through DHCP option 43, the standard CAPWAP controller DNS name, broadcast or a primed controller IP, so I verify one of these. Then I check UDP 5246 and 5247 are not blocked, the AP's clock and certificate are valid, the country code matches, and the WLC has AP licences available.

I would start with a source of truth for devices and variables, then use Ansible playbooks with Jinja2 templates to generate and push standard configs over SSH, keeping them in Git with peer review. A controller such as Catalyst Center could handle day-0 provisioning and compliance. I would start with read-only tasks like backups and audits, then move to small, tested change automation.

First stop the job, then recover through out-of-band console or a management network, using configuration archive rollback or a saved backup. In future, run changes with commit confirmed or reload in as a safety timer, test in a lab, use a dry run or check mode, deploy to one canary switch, then in small batches with automatic post-checks. Store templates in Git with peer review so every change is visible and reversible.