Every question on one page: open one, choose an answer, and see why at once. Signed in, your latest answer to each question counts, and you can answer again after 24 hours.
Curated: · Written: · Reviewed:
Reviewed 100Review pending 2
Answers to reviewed questions count toward this role and the overall leaderboard; your latest answer to each one is the one that counts.
MCQ-39An interface is configured 10.20.30.140/26. A second interface must sit on the same subnet. Which address can it use?⌄
MCQ-82R1's routing table holds three entries covering the 10.4.0.0/16 space: 10.4.1.0/24 from eBGP (AD 20, next hop 192.0.2.1), 10.4.1.0/24 from OSPF (AD 110, next hop 192.0.2.3), and 10.0.0.0/16 from OSPF (AD 110, next hop 192.0.2.2). An engineer withdraws the eBGP /24, reasoning that OSPF's higher administrative distance makes the /16 unusable anyway. With the eBGP route gone, which next hop does R1 use for a packet to 10.4.1.9?⌄
MCQ-83A team adds a /32 static route for 10.4.7.9 via next hop 192.0.2.1, reachable over a link that has flapped twice this month. The default route 0.0.0.0/0 via 198.51.100.1 is healthy. During the last flap the /32 stayed installed with an unresolved next hop for 55 minutes and every packet to 10.4.7.9 was dropped. Which check actually proves the design will fail over to the default?⌄
MCQ-85A router's line card programs IPv4 routes into a 512,000-entry TCAM. After a full-table BGP session comes up, the peer is Established and the RIB holds 518,000 IPv4 prefixes. Monitoring reports the hardware FIB holds 512,000 entries. Pings to several /24s in the newly learned set fail, while pings to the peer's loopback and to the default route succeed. Which explanation fits the evidence?⌄
MCQ-23R1 originates 10.8.0.0/16 to its upstream and installs a static discard route for it. It learns four /24s from R2: 10.8.1.0/24, 10.8.2.0/24, 10.8.3.0/24 and 10.8.4.0/24, and is configured to originate the /16 whenever at least one of them is present. R2's link to the 10.8.4.0/24 site fails and that /24 is withdrawn; the other three stay up. Upstream routers keep 10.8.0.0/16 pointing at R1. Pings to 10.8.1.10 succeed; pings to 10.8.4.10 fail for 1 hour 15 minutes. Which statement explains the failure?⌄
MCQ-74Host A (10.10.0.11/24) sits in VLAN 10 and host B (10.20.0.22/24) in VLAN 20 on the same access switch. Inter-VLAN traffic is routed by a router-on-a-stick: one physical router interface with two subinterfaces, encapsulation dot1Q 10 and encapsulation dot1Q 20, hanging off an 802.1Q trunk whose native VLAN is 1 on both ends and whose allowed list is 10, 20. Which description of the tagging and forwarding is correct?⌄
MCQ-73R1, a Cisco IOS-XE edge router in AS 65500, learns 203.0.113.0/24 from two directly connected eBGP peers. From 192.0.2.9 (AS 65001): weight 200, local preference 100, AS_PATH 65001 65002 64500, MED 50. From 198.51.100.9 (AS 65003): weight 100, local preference 200, AS_PATH 65003 64500, MED 10. The local preference on the second route was set by an inbound route-map, `bgp always-compare-med` is unset, and no other best-path knob has been changed from default. Which path does R1 install, and why?⌄
MCQ-6SW1 and SW2 are joined by an 802.1Q trunk that is up and passing tagged traffic for VLANs 10 and 20, both of which are in the trunk's allowed list. Both switches run PVST+ and neither has a DHCP relay configured. SW1's native VLAN is 10; SW2's native VLAN is 20. A host in VLAN 20 on SW2 receives a DHCP lease from the VLAN 10 subnet, and `show spanning-tree vlan 20` on SW2 names SW1's VLAN 10 root bridge as the root for VLAN 20. What explains both symptoms?⌄
MCQ-3At 14:02 an alert fires on VLAN 40. Over the next 4 minutes the access switch's 1.2 Gbps uplink to the distribution switch reaches 98% utilisation, all of it unknown-unicast and broadcast. A traceroute from a VLAN 40 host to its default gateway (the access switch SVI) completes in 1 hop in 3 ms. The distribution router's ICMP time-exceeded counter for the same interval reads 0. The two access switches' MAC move counters each increment about 40,000 times per minute, for the same 12 MAC addresses. Which action addresses the fault?⌄
MCQ-4All 40 switches in a campus run the default STP priority 32768. A newly racked access switch, ACC-9, has the lowest bridge ID in the topology. Since it was added, east-west traffic between two servers on different distribution switches takes 11 hops instead of 2, and latency rose from 6 ms to 41 ms. Which change restores the intended forwarding path?⌄
MCQ-8A four-member LACP port-channel passed its change validation with every member up/up and collecting-distributing, yet for 90 minutes 1 in 4 application sessions failed: one member was dropping payload while its LACP state stayed healthy. Which validation step would have caught the fault before the change closed?⌄
MCQ-87SW1 is a 24-port access switch with PortFast on every access port and two uplinks (Gi1/0/1, Gi1/0/2) to the distribution layer, which holds the STP root. BPDU guard is enabled on all 26 ports. After a distribution switch reloads, both uplinks go err-disabled within seconds of coming up and stay down, errdisable recovery is not configured, and the 40 users behind SW1 lose connectivity. Which change restores the uplinks without leaving the access ports unprotected?⌄
MCQ-88An access switch's uplink Gi1/0/24 is configured with PortFast and connects to a distribution switch over an 802.1Q trunk carrying VLANs 10, 20 and 30. On link-up the port goes straight to forwarding; for about 30 seconds broadcast frames flood every access port at line rate and the switch CPU pegs at 100%, until STP blocks the port. Which change addresses the cause?⌄
MCQ-86A 24-port access switch carries VLAN 20 on a 1 Gbps uplink to distribution. STP is disabled on VLAN 20. Storm control on the uplink is set to 1 percent for broadcast and its drop counter is incrementing. Users report loss; access ports offer 240 Mbps of broadcast traffic. The MAC address table shows the same MAC moving between Gi0/7 and Gi0/12 about 45 times per second. What should the engineer do or conclude?⌄
MCQ-7A switch logs MAC 00:1a:2b:3c:4d:5e moving between Gi1/0/12 and Gi1/0/18 roughly 4,200 times per minute on each port, sustained for 22 minutes. Both ports are members of Port-channel 10 and carry VLAN 30 only. No other MAC in VLAN 30 moves, spanning tree recorded no topology change in the last hour, and the VLAN's broadcast rate holds at 1.2 Mbps while unknown-unicast to that address is flooded. The first ticket against it was filed as a gateway ARP fault. What do these readings establish?⌄
MCQ-5A change adds VLAN 47 to the allowed list on a LAG trunk between two access switches. The far switch is a different vendor's platform where the allowed list is configured separately and `show interfaces trunk` is not available. The ticket says 'interface is up, so VLAN 47 now passes.' Which statement should the review accept?⌄
MCQ-15Two routers on the same VLAN have formed an OSPF neighbor relationship, but `show ip ospf neighbor` has shown the peer in EXSTART for 20 minutes. Both interfaces are in area 0 with hello 10 / dead 40, both use network type broadcast, neither uses authentication, and the router IDs are unique. Which change resolves the adjacency?⌄
MCQ-9Core-1 and Core-2 are OSPF neighbors in area 0, and `show ip ospf neighbor` on Core-2 has shown FULL for 5 hours. Core-1 holds 64 static routes and runs `redistribute static` under `router ospf`. None of the 64 appear in Core-2's RIB, and `show ip ospf database external` on Core-2 lists 0 Type-5 LSAs. What does this establish?⌄
MCQ-10Two sites run OSPF: site A in area 1, site B in area 2, joined by a direct link that sits in area 1. Each site has a single uplink into area 0. A core failure takes down both uplinks; intra-area adjacencies stay FULL but inter-area prefixes disappear for 40 minutes. The sites must stay in separate areas for LSA scale, and no new circuits can be provisioned this quarter. Which change prevents the next occurrence of this outage?⌄
MCQ-17R1 and R2 are the two ABRs between OSPF area 0 and EIGRP AS 100. Both run mutual redistribution with no tags or filters, and the OSPF LSDB has grown to 2,900 type-5 LSAs for 180 real prefixes. An engineer changes R1 only: its EIGRP→OSPF redistribution now sets tag 100, and its OSPF→EIGRP redistribution now denies tag 100. R2 is left unchanged. What happens to a prefix that originates in the EIGRP domain?⌄
MCQ-84A router has 800 OSPF neighbors, each sending one 64-byte hello every 10 s; the dead interval is 40 s and control-plane CPU peaks at 35%. Its CoPP policy polices the OSPF class to 20 pps. Every OSPF interface shows zero input drops, zero output drops and zero errors, and data-plane forwarding never stops, yet adjacencies across the router flap every few minutes. Which diagnosis and change fit?⌄
MCQ-22A leaf-spine fabric runs BFD at 50 ms × 3 on its BGP sessions. During a nightly backup window the route processor's control-plane CPU reaches 88%, and one BGP session flaps 14 times in 2 hours. Over the same window the link carries user traffic with 0% packet loss, and its optical receive power, CRC and interface error counters are unchanged from baseline. What is the right response?⌄
MCQ-11R1 (AS 65001) and R2 (AS 65002) are directly connected eBGP peers. The session has been Established for 2 hours with no resets, and the same session also carries the IPv6 unicast family, which is exchanging 340 prefixes. R1's IPv4 unicast BGP table holds 1,140 best paths, the only outbound policy on the R2 neighbour is a single permit sequence with no match clauses and no prefix filters, and R2, which is configured for IPv4 unicast towards R1, receives 0 prefixes from it. On R1, the advertised-routes view for the R2 neighbour is empty. Which explanation fits this state?⌄
MCQ-12A team adds a new eBGP session to a transit provider on a router whose BGP stack implements RFC 8212. The neighbor is configured with the peer's AS number and address only. Twenty minutes later `show bgp summary` reports the neighbor Established with 20:14 uptime and a PfxRcd count of 0. The provider's looking-glass shows the session Established, about 900,000 prefixes advertised toward the customer, and none of the customer's 40 prefixes received; their NOC confirms no outbound filter on their side. The session was already cleared once and the count returned to 0 within a minute. What is the correct next step?⌄
MCQ-13AS 65001 has two links to the same transit provider, AS 65010: a primary and a backup. The team wants all outbound traffic to leave via the primary. Their change adds a route-map that sets MED 50 on advertisements sent to AS 65010 over the primary link, and leaves local preference at its default on both links. Outbound traffic still leaves via the backup. Which statement should the review accept?⌄
MCQ-14A voice platform in AS 64500 sends RTP to a partner prefix in AS 65010. The edge routers run default BGP best-path selection. Two egress routes sit in the RIB: via transit AS 3356, AS_PATH length 2, RTT 91 ms, 0.6% loss; via a regional route server, AS_PATH length 4, RTT 12 ms, 0.0% loss. The transit route is installed and the voice cluster has ridden it for three days. Which change puts RTP on the 12 ms path?⌄
MCQ-16We are AS 65001. Customer AS 64512 reaches us over one 1 Gbps eBGP session; transit providers AS 174 and AS 3356 each have their own session. At 14:02 the customer session advertises 0.0.0.0/1 and 128.0.0.0/1 with AS_PATH "64512 174 3356". Both install in the RIB, and because they are longer than every /8–/24 we hold, longest-prefix match sends all outbound traffic to the customer circuit, which saturates at 2.1 Gbps offered and drops. The customer session stays Established; the transit sessions never flap. Which single change, in place before 14:02, would have kept those two prefixes out of the RIB?⌄
MCQ-20A regional ISP's edge router is dual-homed to two transit providers. Legitimate inbound packets from provider A's customers carry source prefixes the router learns only via provider B, and their return path follows B. The team wants uRPF enabled on both provider-facing interfaces without dropping that traffic. Which configuration is defensible?⌄
MCQ-102An operator's RPKI policy rejects Invalid routes and accepts Valid and NotFound. Two IPv4 announcements arrive: 203.0.113.0/24 with origin AS 64510, where the covering ROA authorizes 203.0.113.0/24 for AS 64520; and 198.51.100.0/25 with origin AS 64530, where the covering ROA authorizes 198.51.100.0/24 for AS 64530 with maxLength 24. Which routes does the policy reject?⌄
MCQ-2A PE router runs eBGP with a peer at 10.0.0.1 on a directly connected subnet. The RIB holds 14,112 selected IPv4 prefixes; the line card's hardware FIB holds 12,004, and the FIB reports 12,004 of 64,000 entries in use. The 2,108 prefixes in the RIB but not the FIB all recurse through next-hop 10.0.0.1, while prefixes through 10.0.0.2 and 10.0.0.3 are all programmed. `show ip route 10.0.0.1` shows the next-hop resolved, `show ip arp 10.0.0.1` shows a complete entry, and a hardware exact-route lookup for a destination in one missing prefix returns no forwarding entry. What explains the 2,108-prefix gap?⌄
MCQ-18A core refresh was signed off after one ping to a host behind it; the ping replied. Eight minutes into production traffic, a large share of flows is being dropped even though that host still answers a ping. What should the team do?⌄
MCQ-19Two firewalls, FW-A and FW-B, sit in an ECMP pair in front of a server farm and keep separate state tables. After a third equal-cost path was added, new TCP connections to the farm fall from 3,000/min to 0. Over the next 8 minutes FW-A logs 0 drops, FW-B logs 26,000 drops all classified 'out-of-state TCP', both firewalls sit at 18% CPU, and FW-B's state table is 12% full. A capture at FW-A shows the client's SYN arriving and being forwarded; a capture at FW-B shows the server's SYN-ACK arriving with no matching session. Which statement explains the failure?⌄
MCQ-21Two firewalls run VRRP: FW-A is master, FW-B backup, advertising every 1 s, with no session synchronization configured. FW-A holds 12,600 NAT translations for outbound TCP flows when its power supply fails. FW-B becomes master 3 s later. Hosts on the client subnet receive ICMP replies from the VIP within 3 s, but every established TCP session through the firewall stalls and clients spend the next 12 minutes reconnecting. Which explanation fits these observations?⌄
MCQ-24A VXLAN EVPN fabric has four leaves and two spines; every underlay link is MTU 1500 and no jumbo frames are configured anywhere. Hosts on different leaves complete SSH logins and small pings, but a 200 MB TCP file copy stalls after a few hundred kilobytes and times out. EVPN sessions between all leaves are Established and the underlay routing table is fully converged. Which change addresses the cause?⌄
MCQ-25Leaf-1 (loopback 10.255.0.1) and leaf-2 (loopback 10.255.0.2) are VTEPs in one VXLAN EVPN fabric over an OSPF underlay. Host 10.0.0.11 sits on leaf-1 and host 10.0.0.12 on leaf-2, in the same /24; the tenant's VLAN 10 is designed to be VNI 10100 on both leafs. BGP EVPN has been Established for 45 minutes and each leaf holds the other's Type-2 route for the tenant MAC. A continuous ping from 10.0.0.12 to 10.0.0.11 with a 64-byte payload never gets a reply. Counters read at minute 45:
leaf-1: NVE peer 10.255.0.2 Up; local VNI 10100 Up; decapsulated packets for VNI 10100: 0; outer UDP/4789 packets received from 10.255.0.2: 12,004; uplink input drops and errors: 0.
leaf-2: encapsulated packets for VNI 10100: 0; encapsulated packets for VNI 10101: 12,004.
Which explanation fits these counters?⌄
MCQ-26A VXLAN EVPN fabric carries tenant traffic between leaf-1 and leaf-4 over eight equal-cost underlay paths. A probe sends 100 large (1400-byte) packets on each of 800 flows, with 100 flows pinned to each path by its 5-tuple. Exactly 100 flows lose all 100 packets; the other 700 lose none. The same flows lose nothing with 64-byte packets. BGP EVPN is Established and the FIB on leaf-1 lists all eight next hops. Which explanation fits, and what is the right next step?⌄
MCQ-27A provider runs MPLS L3VPN, and PE-1 hosts both tenants' VRFs. Tenant A's VRF on PE-1 exports route target 65000:100 and advertises 10.20.0.0/16 plus a /32 host route for its billing database at 10.20.7.7. Tenant B's VRF on PE-2 imports only 65000:200. During a change window an engineer adds 65000:100 to tenant B's import list on PE-2, and it stays for nine days. Tenant B's app server 10.30.1.5 sends a query to 10.20.7.7. What happens to the query?⌄
MCQ-28A branch router runs a route-based IPsec tunnel to HQ with traffic selectors 10.8.0.0/24 <-> 10.9.0.0/24. IKE phase 2 completed two hours ago and the SA reads up on both peers, yet a capture at HQ shows a monitoring host's 880 MB of traffic to 10.9.0.5 arriving as cleartext. Which check settles whether this SA is carrying that flow?⌄
MCQ-33BR1 is a branch router with a single WAN link, Gi0/0, addressed 198.51.100.2/24 with ISP next hop 198.51.100.1. Tunnel0 is a GRE tunnel, source 198.51.100.2, destination 203.0.113.7 (the hub's public address), and it carries the branch's only path to the hub networks in 10.0.0.0/8. Last night's change removed ip route 0.0.0.0/0 198.51.100.1 and installed ip route 0.0.0.0/0 Tunnel0, leaving the 10.0.0.0/8 route on Tunnel0. Seconds later Tunnel0 reports line protocol down, 10.0.0.0/8 is unreachable, and a ping to 198.51.100.1 still succeeds. Compliance requires that branch internet traffic keep traversing Tunnel0 for inspection at the hub. Which change restores connectivity and meets that requirement?⌄
MCQ-34Branch firewall A (203.0.113.10) and cloud firewall B (198.51.100.20) run an IKEv2 IPsec tunnel. NAT-T is not in use, so ESP is protocol 50. Both firewalls show IKE_SA and CHILD_SA as established; the vendor dashboard shows 'VPN up' green. Over 4 hours, A's ESP outbound counter is 1,240 packets, but B's ESP inbound counter is 0. A TCP flow from 10.0.0.1 to 172.31.0.10 never completes. A capture on B's outside interface shows IKE packets to UDP 500 but no protocol-50 packets. A's inside ACL permits 10.0.0.0/24 to 172.31.0.0/16, and A's routing table has 172.31.0.0/16 via the tunnel interface. What should you do next?⌄
MCQ-35Two routers run a site-to-site IPsec tunnel with ESP and a 64-packet anti-replay window. The underlay hashes each flow across four equal-cost paths. Over 18 minutes the tunnel drops 2,200 inbound ESP packets; the IKE SA stays up, the integrity-failure counter stays at zero, and the ESP drop counter accounts for all 2,200. Reorder depth measured on the underlay peaks at 90 packets. What is the cause and the fix?⌄
MCQ-30A branch gateway at 10.20.30.41 sits behind a port-mapping NAT that translates it to 203.0.113.7. The site has no public address and the appliance supports no static one-to-one rules, so the addressing cannot change. It dials an IKEv2 tunnel to a hub at 203.0.113.10. The firewall policy between the two sites permits UDP 500 in both directions, denies UDP 4500, and has no protocol 50 rule.
IKE_SA_INIT is answered in 20 ms and both gateways log that a NAT is on the path. The branch then retransmits IKE_AUTH five times, at 0.7 s, 1.4 s, 2.8 s, 5.6 s and 11.2 s, and gives up. After 20 minutes neither gateway shows a child SA and the ESP byte counters still read 0. Which change gets the tunnel carrying traffic?⌄
MCQ-31Two data centres are joined by a site-to-site IPsec tunnel (tunnel mode, ESP). The provider link has a fixed 1500-byte MTU and the tunnel encapsulation adds 50 bytes. The tunnel interfaces are left at the default 1500-byte MTU. TCP handshakes and small requests complete, but large file transfers stall after the handshake; the IPsec SA stays up and 64-byte keepalives pass. Traceroute shows the underlay routers' ICMP 'fragmentation needed' messages are dropped by a firewall the team does not control. Which change clears the stalls?⌄
MCQ-77A branch firewall runs a policy-based IPsec tunnel to HQ with crypto ACL 110: permit ip 10.20.0.0 0.0.255.255 10.60.0.0 0.0.255.255, mirrored on the HQ peer. A 10.44.0.0/24 voice subnet was added at the branch last week for the call centre, and calls to the HQ contact centre on 10.60.7.0/24 now connect with one-way audio. A 60-second capture on the branch's WAN interface during a busy period shows 4,218 UDP/RTP packets sourced from 10.44.0.0/24 to 10.60.7.0/24 carried with no IPsec framing (no IP protocol 50, no UDP 4500), alongside 9,900 ESP packets to the HQ peer address. In the same window show crypto ipsec sa lists SAs only for proxy 10.20.0.0/16 - 10.60.0.0/16 with 2.1 GB encapsulated, the IKEv2 SA is 4h12m old, the byte lifetime is 4.6 GB, and the OSPF adjacency to HQ is full. What is the correct conclusion, and what should the team change?⌄
MCQ-78Office A and Office B both use 10.10.0.0/16. A host at A (10.10.1.10) opens a TCP connection to a server at B that also has 10.10.1.10. The IPsec tunnel is up, the SA is established, and both peers have selectors for 10.10.0.0/16. A traceroute from A to 10.10.1.10 shows one hop to the local switch, and 64 connections to 10.10.1.10:80 complete against the local host, not the remote server. Which change will make A's connection reach B's server?⌄
MCQ-79Two tenants share a PE router. VRF-A and VRF-B both carry 10.0.0.0/24. Both VRFs were configured with RD 65000:1; VRF-A imports and exports RT 65000:10, VRF-B imports and exports RT 65000:20. On a remote PE, VRF-A's prefix is present, VRF-B's is missing, and `show bgp vpnv4 unicast` lists exactly one route for 65000:1:10.0.0.0/24, carrying RT 65000:10. What explains the missing route?⌄
MCQ-29Two sites are joined by a GRE tunnel over a shared carrier underlay. The change request calls it a VPN and asks for sign-off that inner traffic is confidential. A capture taken on the underlay shows 4.1 GB of inner payload in the clear over the last day; the GRE tunnel has been up the whole time with no flaps and no IPsec transform configured. What should the review do?⌄
MCQ-75A security review taps the provider handoff between two sites. The link carries a GRE tunnel between the two PE routers; each tenant sits in its own VRF and the tunnel interfaces are up/up. In one afternoon the tap captures 880 MB of tenant traffic with the inner 10.0.0.0/8 headers and HTTP request lines readable in the capture. The provider sells a plain Ethernet handoff and runs no link encryption of its own. Which change makes the tap unable to read tenant payloads?⌄
MCQ-76An engineer mirrors a spine-to-leaf uplink that carries only underlay traffic and captures for 10 minutes. Wireshark decapsulates the VXLAN frames and shows tenant A's inner Ethernet frames for VNI 10010, ARP requests and DHCP offers, in plaintext. The overlay runs VXLAN with no IPsec and no MACsec; VTEPs authenticate each other only with a BGP EVPN password. Which conclusion is correct?⌄
MCQ-56A service runs over a GRE tunnel whose interface MTU is 1400 on a 1500-byte underlay. An ACL on the tunnel path drops all ICMP. TCP handshakes complete, small pings succeed, and traceroute returns, but bulk transfers stall: 1500-byte DF probes are dropped while 1400-byte DF probes pass. Which change addresses the cause?⌄
MCQ-57A site's edge ACL permits ICMP echo request and echo reply and denies every other ICMP type. After it was applied, SSH logins to a host behind a GRE tunnel still succeed, but `scp` of a 50 MB file to that host stalls immediately after authentication, the sender retransmitting the same segment until the connection times out. The tunnel MTU is 1400, the physical link is 1500, `ping -M do -s 1472` to the host fails, `ping -s 100` succeeds, and interface counters on both tunnel endpoints show no errors or drops. What is the cause?⌄
MCQ-99A branch router has two egress paths. The IPsec tunnel to HQ has interface MTU 1400 and `ip tcp adjust-mss 1360` configured; bulk transfers over it complete normally. The direct internet interface has MTU 1500 and no MSS clamp, and a carrier middlebox on that path silently drops packets larger than 1400 bytes without returning ICMP Fragmentation Needed. Bulk uploads from branch clients to a SaaS endpoint over the direct internet path stall after the first few KB while the handshake and small requests succeed; 18% of large transfers fail. What should the engineer do?⌄
MCQ-80A 4,200-person company runs always-on VPN as split tunnel against the gateways 198.51.100.10 and 198.51.100.11, with 10.20.0.0/16 and 172.16.0.0/12 in the client include list. One laptop on hotel Wi-Fi lost its IKE SA for 12 minutes 4 seconds; the client UI read Reconnecting the whole time. That laptop's Wi-Fi capture over the outage holds 3,914 packets, 2.3 MB, addressed to 10.20.0.0/16 - 47 DNS queries to 10.20.0.53 and 214 SMB SYNs to 10.20.8.11:445 - plus 9.1 MB to public destinations, none of it encrypted. Security now requires that while an SA is down, nothing but IKE/IPsec to the two gateways leaves a laptop's physical adapter. Which change puts that in place?⌄
MCQ-32A 4,000-laptop remote fleet runs split tunnel. The exclusion list is three summaries: 10.0.0.0/8, 172.16.0.0/12 and 192.168.0.0/16, about 17.9 million addresses, while only six subnets, roughly 1,100 addresses, need the local path. A hunt found malware reaching three addresses off-tunnel for nine days: 10.44.7.19, 10.60.12.4 and 192.168.1.77. None of the three shows up in the egress firewall logs, which see only tunneled traffic, and the pcaps show all three were contacted by literal IP. 1,200 of the 4,000 devices are unmanaged BYOD with no firewall agent, and the concentrator peaks at 1.9 Gbps of its 2 Gbps capacity while remote users push 12 Gbps at peak. Which change closes all three off-tunnel paths while keeping the six local subnets local and the concentrator within capacity?⌄
MCQ-93A Juniper router at a site 2 hours from the nearest console has no out-of-band management. An engineer applies a firewall filter change to the management interface with `commit confirmed 10` and immediately loses the SSH session, because the filter drops traffic from the engineer's own subnet. No one else can reach the device. What happens next?⌄
MCQ-47example.com is delegated from .com to two NS records, ns1.example.com and ns2.example.com. ns2 was repurposed last month as a web front end and nothing now listens on its port 53. Monitoring shows 35% of queries for example.com return SERVFAIL while the rest resolve normally. Which action fixes the delegation?⌄
MCQ-45A zone's SOA has MINIMUM 900 and an SOA record TTL of 3600. At 13:50 a recursive resolver caches NXDOMAIN for api.example.com. At 14:00 the team creates the A record for api.example.com (TTL 300) on the authoritative server, which answers correctly from then on. When does that resolver first return the A record?⌄
MCQ-44A service's A record has TTL 3600 s. At 09:00 the team edits the record to the new address 203.0.113.20 and lowers the TTL to 60 s in the same change. By 09:30 monitoring shows 38% of requests still arriving at the old address 203.0.113.10. The next cutover must move all traffic within 5 minutes. What should the team do?⌄
MCQ-46At 09:00 you change the A record for api.example.com on ns1.example.com from 10.0.1.5 to 10.0.2.7; the zone's TTL is 3600 s. At 09:20 a service on the corporate network still connects to 10.0.1.5. `dig @ns1.example.com api.example.com` returns 10.0.2.7 with TTL 3600. `dig @10.20.0.53 api.example.com`, where 10.20.0.53 is the resolver the corporate host is configured to use, returns 10.0.1.5 with TTL 2400. What is the correct next step?⌄
MCQ-48`api.example.com` holds a TXT record that a vendor's domain-verification check reads at exactly that name. The team wants the same name to point at a load balancer whose hostname is `api.lb.net`, so they add `api.example.com. 300 IN CNAME api.lb.net.` next to the TXT and reload the primary (BIND 9.18). The reload fails with `CNAME and other data`, and the primary keeps serving the previous zone, so `api.example.com` still answers with its old A record. The vendor will not accept the TXT at any other name, and the ~200 clients are configured with `api.example.com` and cannot be repointed. What is the correct action?⌄
MCQ-49A company signs corp.example and publishes its DS record at the registrar. Its clients are switched to DoH pointed at a public resolver. A red-team test shows clients accepting a forged A record for corp.example, and the DoH TLS session to that resolver was never broken. Which change makes clients reject the forged record?⌄
MCQ-70A service uses geo-DNS steering with a 900 s TTL on its steering records. At 12:00 the team removes PoP-B from the geo map. At 12:03, monitoring shows 40% of application requests still arriving at PoP-B. What explains the residual traffic?⌄
MCQ-50A printer with a DHCP reservation for 10.20.20.8 goes offline for 40 minutes. The DHCP server log shows a normal ACK for the reservation, and the switch has learned the printer's MAC on a different access port. Which change prevents the recurrence?⌄
MCQ-51A relay agent on a Layer 3 access switch relays DHCPv4 for client subnet 10.20.0.0/21; giaddr is 10.20.0.1. Two access ports share that subnet: Gi1/0/1 serves 400 IP phones that need option 66 = 10.20.0.50 to reach their TFTP server, and Gi1/0/2 serves 600 workstations that must not receive option 66. The DHCP server has one scope for 10.20.0.0/21 whose default class sets no option 66, plus a phones class that sets option 66 = 10.20.0.50 and matches option 82 remote-id 'Gi1/0/1'. The relay is configured to insert option 82 with the circuit-id sub-option only. All 400 phones now receive no option 66 and fail to reach the TFTP server; the 600 workstations are unaffected. What is the correct fix?⌄
MCQ-98A conference venue's guest Wi-Fi runs one DHCP scope, 10.50.0.0/24 with 4 addresses excluded for infrastructure (250 usable), on an 8-hour lease. Attendees join at about 200 an hour from 09:00, connect once, and stay a median of 40 minutes. At 10:15 the server logs that it has nothing left to offer, phones in the lobby get no address, and the wireless controller shows 133 clients associated. Renumbering the VLAN to widen the scope needs a change window after the event. What restores new connections today?⌄
MCQ-52A dual-stack VLAN runs stateful DHCPv6 (M flag set) for addresses and DNS, and the router is configured not to send Router Advertisements, to keep the segment quiet. Hosts get IPv6 addresses and resolve AAAA names, but `ping6` to an off-link address fails while IPv4 to the same destination works; `ip -6 route show` lists the on-link prefix and no default. What is the correct fix?⌄
MCQ-55A VLAN ACL on the aggregation switch is `permit ethertype 0x0806 any any` followed by an implicit deny. On the 300 dual-stack hosts in that VLAN, `ping 10.40.0.1` succeeds and `arp -a` lists the gateway, but `ping6 fe80::1` fails and `ip -6 neigh show` lists the gateway as FAILED. Which conclusion and fix are correct?⌄
MCQ-54R1 runs IOS XE 17.9 with SVI VLAN 20 at 198.51.100.1/24 and a static route 10.20.0.0/16 via 198.51.100.50. At 09:12, monitoring shows 3,200 drops per minute to 10.20.0.0/16. `show ip arp 198.51.100.50` gives Incomplete, age 00:02:40. A capture on VLAN 20 shows R1's who-has 198.51.100.50 leaving every second and no reply arriving. A workstation 198.51.100.60 on VLAN 20 pings 198.51.100.50 in 1 ms. R1 pings its own 198.51.100.1 but not 198.51.100.50. VLAN 20 has a VACL with counters. Which action best identifies and fixes the cause of the drops?⌄
MCQ-53A payments API is dual-stack. Its A records point at a load balancer whose IPv4 path works, with 60 ms p99 for the checkout handshake (about a 30 ms round trip, so TCP plus a one-round-trip TLS setup finishes roughly 60 ms after the IPv4 SYN is sent). A DNS change added AAAA records, and a firewall rule now drops every packet to that IPv6 address silently, with no RST and no ICMP. Ten minutes in, p99 for checkout is 310 ms on clients that prefer IPv6, and about 4 percent of sessions, an older mobile SDK that opens only an IPv6 connection when AAAA is present, fail after about 127 s, the Linux connect budget with tcp_syn_retries=6. Client stacks use a Connection Attempt Delay of 250 ms. The IPv6 load balancer's listener reports up, and its health checks run from the IPv4 management network and stay green. Which action should the review require before this change is kept?⌄
MCQ-36A service runs four identical binaries behind a round-robin load balancer. The hosts differ in size: one has 2 vCPUs, two have 8 vCPUs, and one has 32 vCPUs. Every request is CPU-bound and costs about 1 ms of CPU. Traffic is steady at 8,000 requests/s. Within a minute the 2-vCPU host's p99 climbs from 45 ms to 900 ms with CPU pinned at 100%, the two 8-vCPU hosts sit at 25% CPU each, and the 32-vCPU host reports 6%. What do these numbers show, and what change fixes it?⌄
MCQ-37A change proposes keeping least-connections balancing on the L4 balancer after clients migrate to HTTP/2. In the 26-minute window after the migration, each of the four backends holds exactly one TCP connection, but stream counts are A 800, B 70, C 70, D 60. Which statement should the review accept?⌄
MCQ-38A pool of six members sits behind a load balancer whose health check is a TCP connect to port 8443 every 5 s, ejecting a member after 3 consecutive failures. A deploy ships a build whose HTTP handler returns 503 for every request while the TLS listener still completes handshakes. Over the next 11 minutes the pool serves 7,200 requests, all 503, and the balancer ejects 0 members. Which change ejects the members during that failure?⌄
MCQ-41A checkout service runs behind an L4 load balancer. During a deploy a member is pulled from the pool while it holds 940 open connections: 910 are short calls that finish in under 2 seconds, and 30 are checkouts that take 45 seconds end to end and roll back if the connection is reset. Connection draining is enabled with a 10-second timeout. What does the removal produce?⌄
MCQ-43A web tier runs four identical members behind a load balancer that inserts a session cookie with a 30-minute TTL; each member keeps sessions in its own heap. Health checks mark a member down 5 seconds after it stops answering and remove it from rotation. During a rolling restart that takes each member down for 40 seconds, one at a time, about 25% of requests return 502 from the balancer, and those 502s continue for up to 30 minutes after the last member is healthy again. Requests arriving without a session cookie succeed throughout. What explains the failures?⌄
MCQ-40A checkout service calls a payments dependency through three layers (edge, mid, data). Each layer is configured with "3 retries" and a 1 s timeout. During an 8-minute brownout of the dependency, checkout p99 rises from 300 ms to 12 s and the dependency reports 27× its baseline request rate. Which change addresses the cause?⌄
MCQ-69A payments API is announced as one anycast VIP from three PoPs. Sessions are long-lived, median 22 minutes. During a maintenance window the team lowered IGP cost on the Dublin uplink; within 30 seconds 9% of established sessions were reset with RST, and every reset session was one whose flow rehashed to Dublin. Requests shorter than 2 seconds saw no errors. The session-drop SLO is 0.5%, and each reset forces a client retry that can double-charge. The team has one sprint and cannot deploy cross-PoP TCP state replication or change client behavior. What should they do?⌄
MCQ-95A DNS service is announced as the same /24 from four anycast PoPs. One PoP's local resolver fails and answers SERVFAIL, but its interface and BGP session stay up, so the prefix stays advertised. Synthetic monitoring runs from three vantage points that all hash to healthy PoPs and reports 100% success, while 22% of production clients, those BGP-steered to the failed PoP, get SERVFAIL for 19 minutes. Which action stops the SERVFAILs for those clients?⌄
MCQ-63An HTTPS load balancer terminates TLS on 203.0.113.10 for 30 hostnames, each with its own certificate, and selects the certificate from the SNI value the client sends. A monitoring system runs a TLS check against https://203.0.113.10/ with no SNI and fails with `x509: certificate is valid for www.example.com, not 203.0.113.10`. All 30 hostname-based clients connect successfully. What explains the failure and what should the monitor do?⌄
MCQ-71An ALB terminates TLS on port 443 with an ACM certificate and forwards to a target group configured for HTTP on port 8080. A capture on the backend subnet shows 2.4 GB of cleartext HTTP over three days. Which statement about this setup is correct?⌄
MCQ-72From one client subnet, ICMP echo to 10.20.30.140 answers in 0.4 ms and the application's UDP channel to the same host (port 9999) works in 6 ms. TCP connections to 10.20.30.140:8443 from that subnet send SYN, retransmit six times over about 127 seconds, and time out with no RST and no SYN-ACK. Which diagnosis and next step are defensible?⌄
MCQ-58A service intermittently times out. tcpdump on a client shows 8,400 SYNs to the service port with no SYN-ACK returned. On the server, the listen queue sits at 3 of 128, the conntrack table is at 12% of its maximum, and the host has one NIC and one default route. The application log records no inbound connection attempts. What is the correct next step?⌄
MCQ-59Connections from a client to a service are reset mid-transfer. A capture at the client shows each RST arriving with TTL 61, while the service's data packets arrive with TTL 57. A capture on the service host's NIC shows no RST leaving the host, and the service's logs show the connections still open when the client receives the RSTs. The RSTs carry sequence numbers matching the current connections. What does this establish?⌄
MCQ-60An engineer investigating a slow 2 GB backup copy captures on host A and sees TCP segments with 65,495-byte payloads, and suspects a jumbo-frame or MSS problem. A SPAN capture of the same flow on the access switch shows 1514-byte frames carrying 1460-byte TCP payloads, with no drops and no retransmissions. The path MTU is 1500 end to end. Which conclusion is correct?⌄
MCQ-61Host H1 in 10.4.0.0/24 intermittently fails HTTP requests to a partner API. The change under review proposes a 90-minute routing change to remove a suspected black hole at hop 6, based on this evidence: `traceroute` (UDP, default port 33434) from H1 stops at hop 6, while a packet capture on H2 in 10.9.0.0/24 shows a completed TCP 443 transaction with the same partner. Which statement should the review accept?⌄
MCQ-62A branch-to-data-center TCP flow reports 1.2% retransmissions and 40 ms RTT. During a 10-minute loss window, engineers run captures and drop counters at three candidate hops. The access switch uplink output-drop counter increments by 1,200. The WAN edge and data-center edge counters do not increment. What is the correct conclusion?⌄
MCQ-64A 10G uplink Te0/0/0 on router R1 was shut down during maintenance. After the shutdown is removed, `show interfaces Te0/0/0` reports `up`, `line protocol is down`, and `0 packets/sec` input and output for 25 minutes. OSPF adjacency does not form; the NOC restarts the OSPF process twice. `show interfaces Te0/0/0 transceiver` reports Rx power `-40.0 dBm`. Which action addresses the cause?⌄
MCQ-65Policy requires every packet between the client subnet 10.20.1.0/24 and the payments subnet 10.30.7.0/24 to transit the stateful firewall FW1, which is the payments subnet's only gateway. Last night an operator added a route on R1, the client-side router: 10.30.7.0/24 now goes to R2, a router sitting on the payments segment, instead of to FW1. Since that change, every connect from 10.20.1.10 to 10.30.7.20:8443 fails: in 12 minutes the client sent 40 SYNs and received no SYN-ACK, a capture on 10.30.7.20 shows 40 SYNs in and 40 SYN-ACKs out, and FW1's counter for dropped return-path packets rose by exactly 40. FW1's configuration is unchanged. Which change restores connectivity and keeps the policy satisfied?⌄
MCQ-66A nightly backup copies data from a server in Frankfurt to a client in Singapore over a single TCP connection. The transfer stalls at about 0.6 Mbps. A capture on the client shows the receiver's advertised window (rwnd) at 2920 bytes, the sender's congestion window (cwnd) at 64 segments, no retransmissions, and no ECN marks. RTT is 40 ms. What is the correct remediation?⌄
MCQ-67A metrics agent sends UDP datagrams at 200/s to 10.4.0.7:8125. A restart points the collector at 8126 instead, so nothing owns 8125 for 11 minutes; the host stays up and its kernel answers the flow with ICMP destination unreachable, port unreachable (type 3, code 3). Over that window the agent's send counter holds at 200/s, its application-level error counter reads 0, and the collector's receive counter reads 0. The agent's socket is unconnected and the code never reads the socket error queue. The vendor's wire format is UDP-only, so switching transports is off the table for now. Which conclusion and immediate action fit the evidence?⌄
MCQ-68A receive-only RTP client sits behind a CGNAT that allocates a separate external port per destination (address-and-port-dependent mapping), refreshes a mapping only on outbound packets, and expires an idle UDP mapping after 30 s. At setup the client sent one packet to 203.0.113.9:5004 and the sender learned the external port it created; since then the client's only outbound traffic is a 20-byte keepalive every 20 s to the signaling server at 203.0.113.9:5060. The feed goes silent 30 s into every session while the keepalives keep flowing. What fixes it?⌄
MCQ-96Two racks are joined by a provider handoff that carries 1500-byte frames and cannot be reconfigured. NFS clients in rack A and the NFS server in rack B both have their interfaces set to MTU 9000. A firewall between the racks permits ICMP echo request and reply but drops every other ICMP type. A 64-byte ping to the server succeeds; `ping -M do -s 8972` fails; large NFS reads stall for 30 s and then fail, while small metadata operations succeed. Which change addresses the cause?⌄
MCQ-100Leaf L1 and spine S1 are joined by a 4-member ECMP bundle (m1–m4). A client at 10.20.1.7 opens TCP 443 connections to 10.60.4.20 with 1400-byte payloads and DF set; about one in four connections stalls after the handshake, and a retry always succeeds. One-way probes pinned to each member by source-port sweep (1400-byte DF datagrams, counted at the far end) give:
| Direction | m1 | m2 | m3 | m4 |
| --- | --- | --- | --- | --- |
| client→server | 0% loss | 0% loss | 0% loss | 0% loss |
| server→client | 0% loss | 0% loss | 0% loss | 100% loss |
Which action is defensible?⌄
MCQ-89R1 is the branch WAN edge and marks voice for a provider SLA that classifies on DSCP; the campus QoS configuration is frozen for this window. R1 carries 12 concurrent G.711 calls to HQ (20 ms packetization, so 50 packets per second per direction per call) over a direct 100 Mbps provider Ethernet handoff on Gi0/0. Over 3 hours, a capture taken on R1's side of the handoff shows 6,480,000 RTP packets leaving R1 with DSCP CS0, and the class VOICE entry of the WAN-QOS policy - applied as `service-policy input GigabitEthernet0/0`, matching UDP 16384-32767, action `police 200000` - also reports 6,480,000 packets. Which change makes the outbound calls leave R1 marked EF?⌄
MCQ-101On a congested 100 Mbps WAN interface, IOS-XE applies: class VOICE / priority percent 20; class VIDEO / bandwidth remaining percent 40; class BUSINESS / bandwidth remaining percent 30; class class-default / fair-queue. EF voice is offered 26 Mbps, AF41 video 30 Mbps, AF31 business 10 Mbps, and AF11 bulk 50 Mbps. What does the router forward?⌄
MCQ-90A 10 Gbps link is monitored with NetFlow at 1:1024 packet sampling. A customer reports a 5-second burst of 600 packets that were dropped. Querying the collector 10 minutes after the burst returns zero flow records for that 5-second window, and the NOC wants to close the ticket as "no traffic". Which action is correct?⌄
MCQ-97A ToR uplink to a storage array runs at about 5 Gbps during the nightly batch, and a SPAN session mirrors it, offering that 5 Gbps to a destination port that is a 1 Gbps link to a packet analyser. During the batch the uplink's interface counters show 0 output drops and the array reports no TCP retransmits, while the analyser reports that roughly 80 percent of mirrored frames never arrive and pages the on-call for production packet loss. What does the evidence show, and what should be done?⌄
MCQ-91During a change window an engineer adds `deny ip 10.20.0.0 0.0.255.255 any` to the inbound ACL EDGE-IN on a router. Thirty minutes later the 10.20.0.0/16 subnet is unreachable. `show logging` shows no denies for the new rule, but `show ip access-lists EDGE-IN` shows that deny line with 4,200 matches over the same 30 minutes. The ACE was entered without the `log` keyword. What does this evidence show?⌄
MCQ-92A distribution switch filters the management subnet 10.0.0.0/24 in an ACL you are about to extend. The only in-band path to the switch is SSH from a jump host at 10.0.0.5; the last engineer to touch this ACL locked himself out and needed 40 minutes with a crash cart. The change window is 10 minutes. Which approach is defensible?⌄
MCQ-94A dual-stack service is reachable at 203.0.113.10 (IPv4) and 2001:db8:1::10 (IPv6) and listens on TCP/443 and UDP/443. The edge router has an IPv4 ACL denying 10.10.10.0/24 to 203.0.113.10/32; no IPv6 ACL exists. Over 24 hours the service logs 0 requests from 10.10.10.0/24 and 1,400 requests from 2001:db8:10::/64, all TCP/443. Clients are dual-stack and use Happy Eyeballs. Which change enforces the same deny for the IPv6 admin subnet?⌄
MCQ-81R1 and R2 run VRRP group 10 on VLAN 10. R1 is master with priority 150; R2 is backup with priority 100; preempt is enabled; advertisement interval is 1 s. R1 has a BGP default from ISP A; R2 has a BGP default from ISP B. At 14:02 ISP A drops, R1 removes its default route but keeps its VRRP priority and continues advertising. Hosts on VLAN 10 ARP for the VIP successfully and black-hole for 7 minutes until an engineer manually lowers R1's priority. Which change prevents the 7-minute outage?⌄
MCQ-42An API sits behind one reverse proxy. The proxy appends the connection's peer address to whatever X-Forwarded-For the client sent; it never strips or overwrites the client's entries. The API enforces a quota of 600 requests per minute per caller, keyed on the first entry of X-Forwarded-For. Over 20 minutes, one client, TCP peer 203.0.113.77 as the proxy sees it, sent 180,000 requests, all of them were served, the quota never fired, and the access log shows a different caller address on nearly every line. The proxy's configuration is frozen for this quarter. Which change makes the quota key on the actual caller?⌄