10 free sample questions from a bank of 59, with the correct answers and explanations. No signup required — start practising right now.
C is correct. Load balancing an eBGP prefix that arrives from two different AS numbers needs multipath multiple-as under protocols bgp, because Junos only balances equal-cost paths from a single AS by default.
Load balancing an eBGP prefix that arrives from two different AS numbers needs multipath multiple-as under protocols bgp, because Junos only balances equal-cost paths from a single AS by default.
BGP multipath installs equal-cost paths into the forwarding table, but by default it only compares paths whose originating AS is the same. multiple-as relaxes that. An import policy with load-balance per-packet or per-prefix is a separate control.
R2 has 11.11.11.0/24 from R1 (AS 65001) as active and from R3 (AS 65003) as backup. Both paths are eBGP and originate in different autonomous systems, so Junos will not treat them as multipath candidates unless multiple-as is added to the multipath statement under protocols bgp. The exhibit also shows a load-balance per-packet policy that is never attached to the forwarding table, but that is about the hashing action, not about which paths are eligible.
multipath is the default behaviour and only balances paths from the same AS, so it cannot combine the AS 65001 and AS 65003 paths here.If a prefix is learned from two eBGP peers in different AS numbers and you want ECMP, the missing keyword is multiple-as. Multipath on its own is a same-AS feature.
ACD is correct. Three checks on R1: the routes for 203.0.113.128/25 must have several next hops in the RIB, multipath must be enabled under protocols bgp on R1, and a load-balancing export policy must exist under routing-options for the forwarding table.
Three checks on R1: the routes for 203.0.113.128/25 must have several next hops in the RIB, multipath must be enabled under protocols bgp on R1, and a load-balancing export policy must exist under routing-options for the forwarding table.
BGP routes enter the RIB automatically, but the forwarding table is populated separately, so set routing-options forwarding-table export <policy> is what puts multiple next hops into the FIB. A load-balancing import policy is a different thing and is not required once the export policy exists.
The exhibit shows show route forwarding-table matching 203.0.113.128/25 returning eight addresses that all resolve to the single next hop 10.1.1.1, which is the symptom of no load balancing. A: the RIB on R1 must in fact contain multiple next hops for the prefix or nothing downstream can balance it. D: multipath is a receiving-side knob and only R1 is receiving, so it belongs on R1. C: the forwarding table is built from the RIB by the policies exported under routing-options, so the load-balancing action has to be attached there for the extra next hops to be used.
routing-options forwarding-table export policy, so an import policy alone leaves the extra next hops unused.Distinguish the RIB from the forwarding table. Multipath plus routing-options forwarding-table export is the working combination; an import policy changes what is accepted, not what is installed for forwarding.
ACE is correct. Use overload together with timeout 900 -- in the OSPF instance and in the OSPFv3 instance, with the OSPFv3 statement scoped to the ipv4-unicast realm.
Use overload together with timeout 900 -- in the OSPF instance and in the OSPFv3 instance, with the OSPFv3 statement scoped to the ipv4-unicast realm.
set protocols ospf overload [timeout <seconds>] marks the router LSDB bit and makes it wait for the SPF calculation to finish, suppressing LSAs while it is stale. If no timeout is given the state persists until the statement is deleted. Under protocols ospf3 the realm-qualified form realm ipv4-unicast overload timeout 900 covers IPv4 carried in an OSPFv3 process.
The requirement is to reduce the impact automatically after a reboot, which means the overload state must time out on its own. C does that for OSPF and E does it for OSPFv3. A does the same thing for OSPFv3 but constrains it to the IPv4-unicast realm, which is exactly the IPv4 case the manager described, and it is the realm-qualified spelling Junos accepts under protocols ospf3.
set protocols ospf overload with no timeout leaves the router overloaded until an administrator deletes the statement, so nothing recovers automatically after the reboot.set protocols ospf3 overload with no timeout never clears itself.Any overload statement without a timeout value is a manual fix. In these questions the word 'automatically' is the hint that the timeout is mandatory.
AB is correct. Either configure MSDP on each RP, or run anycast PIM and put the rp-set statement on each RP listing the other RPs in the domain.
Either configure MSDP on each RP, or run anycast PIM and put the rp-set statement on each RP listing the other RPs in the domain.
MSDP (Multicast Source Discovery Protocol) runs between RPs and lets each RP learn the (S,G) state its peers have, so sources behind one RP are usable through another. Anycast PIM gives all RPs one address and uses the rp-set statement plus MSDP or PIM to keep the RPs aware of each other's active sources.
Both A and B operate on the RPs. MSDP on each RP is the classic mechanism for sharing active sources between RPs, and anycast PIM with rp-set on each RP is the alternative: every RP advertises the list of RPs in the domain (up to 15) so that a source discovered by one RP is not lost. The rp-set statement belongs on the RPs because they are the peers that need to know about each other.
rp-set on the source DR routers does nothing: the source DR is not an RP and never participates in the RP-to-RP source sharing that anycast PIM relies on.Ask where the protocol runs. MSDP and rp-set both belong on the RPs; anything placed on a source DR or receiver DR is a distractor.
BD is correct. LDP-signalled VPLS is control-bit 0 only, and BGP-signalled VPLS decouples the service from the transport so the pseudowire can ride an RSVP or an LDP LSP.
LDP-signalled VPLS is control-bit 0 only, and BGP-signalled VPLS decouples the service from the transport so the pseudowire can ride an RSVP or an LDP LSP.
VPLS (RFC 4762) carries Ethernet frames over an MPLS LSP using a split-horizon label and a VC label. LDP signalling requires every remote PE to be configured manually, while BGP signalling (RFC 4761) uses BGP extended communities and autodiscovery to build the full mesh of pseudowires. Either way the pseudowire is just an MPLS LSP, so any label distribution protocol can create it.
B is correct because Junos documents LDP-based VPLS as supporting control bit 0 only -- control bit 1 (which blocks locally learned MACs) is not offered for LDP signalling. D is correct because BGP-signalled VPLS advertises the service endpoints over BGP but does not care how the transport LSP is built; the inner and outer labels ride an ordinary MPLS LSP, and LDP, RSVP-TE, BGP-LU and segment routing are all valid transport choices.
Learn the pairing: LDP signalling = manual site configuration, control bit 0. BGP signalling = autodiscovery, transport-agnostic pseudowires.
CD is correct. This is an EVPN Type 2 MAC/IP route learned over iBGP, and the device that advertised it into EVPN is 192.168.101.5.
This is an EVPN Type 2 MAC/IP route learned over iBGP, and the device that advertised it into EVPN is 192.168.101.5.
A Type 2 MAC/IP Advertisement route carries the MAC address, optionally the IP, the Ethernet Tag, the ESI label and the MAC Mobility extended community. In Junos show route extensive the originator of such a route appears as Originator ID, the BGP Protocol next hop, and the RD administrator, while Indirect path forwarding next hops are the resolved tunnel next hops on the local side.
D is right because the route is printed as MAC/IP (2 entries, 1 announced), which is how Junos labels Type 2 EVPN NLRI. C is right because the entry shows Originator ID: 192.168.101.5, Protocol next hop: 192.168.101.5 and an RD administered by 192.168.101.5, so that PE is the one that put the MAC into EVPN. B is wrong on the face of the output: Local AS: 65101 Peer AS: 65101 is identical, so this is an iBGP-learned route.
Indirect path forwarding next hops as Next hop: ... via et-0/0/0.0 and et-0/0/1.0. They are local equal-cost paths toward the originator, not the devices that advertised the MAC.Local AS: 65101 Peer AS: 65101 -- the AS numbers are identical, so the route was learned over iBGP, not eBGP.In EVPN output the advertiser is the Originator ID / Protocol next hop / RD administrator. Everything under Indirect path forwarding next hops is a local tunnel next hop to that originator.
AC is correct. Configure family inet6 labeled-unicast between the PE routers and family inet6 unicast between PE and CE. mpls ipv6-tunneling is already in place.
Configure family inet6 labeled-unicast between the PE routers and family inet6 unicast between PE and CE. mpls ipv6-tunneling is already in place.
mpls ipv6-tunneling lets an MX encapsulate IPv6 packets with an MPLS label stack and an IPv4 header so they can cross an IPv4 core. That only provides the transport -- BGP still needs two address families: a labelled family for the PE-PE session and a plain unicast family for the customer-facing session.
A is correct because the PE-to-PE session is the only place the IPv6 prefix can be advertised with an MPLS label, and without the label the packet cannot be tunnelled across the IPv4 core. C is correct because the CE routers do their own forwarding and must receive ordinary, unlabelled IPv6 unicast routes from the PE. Together they cover both legs of the VPN: labelled inside the provider, unlabelled at the edge.
family inet6 unicast between PEs gives the PE-PE session no MPLS label to push, so the IPv6 traffic cannot be tunnelled through the IPv4 MPLS network even with mpls ipv6-tunneling enabled.Rule of thumb for IPv6-over-IPv4 MPLS: labelled-unicast on the core-facing session, plain unicast on the customer-facing session.
D is correct. OSPFv3 can carry IPv4 routes, but only on interfaces where family inet6 is configured, because the protocol itself runs over IPv6 link-local addresses.
OSPFv3 can carry IPv4 routes, but only on interfaces where family inet6 is configured, because the protocol itself runs over IPv6 link-local addresses.
OSPFv3 (RFC 5340) replaced the IPv4-specific protocol packets of OSPFv2 with generic IPv6 ones and derives the router ID from a link-local address. Junos therefore requires family inet6 on the interface before an OSPFv3 adjacency will form. IPv4 reachability inside an OSPFv3 process is handled by set protocols ospf3 realm ipv4-unicast, which advertises the IPv4 address family as an extended prefix TLV.
OSPFv3's hello and database description packets are IPv6 packets sent to the reserved link-local multicast address ff02::5, and the interface must hold an IPv6 link-local address for that to happen. So even when the point of the process is to redistribute IPv4 prefixes, the interface has to be in the inet6 address family. The other three statements either invert the relationship or overstate it.
family inet6 caveat.set protocols ospf3 realm ipv4-unicast and the native IPv6 realm coexist in the same OSPFv3 routing instance.OSPFv3 needs IPv6 on the wire even when the payload is IPv4. If an adjacency will not come up, check for family inet6 on the interface first.
BC is correct. Type 2 routes distribute MAC (and IP) bindings; Type 3 routes set up the inclusive-multicast tree that carries broadcast, unknown unicast and multicast. Nothing in the question requires segment routing or VXLAN.
Type 2 routes distribute MAC (and IP) bindings; Type 3 routes set up the inclusive-multicast tree that carries broadcast, unknown unicast and multicast. Nothing in the question requires segment routing or VXLAN.
EVPN (RFC 7432) reuses MP-BGP to signal an Ethernet service. Type 2 is MAC/IP Advertisement and is the control-plane replacement for MAC learning. Type 3 is Inclusive Multicast (Ethernet Tag) and is the ingress-replication control plane for BUM traffic. Type 1 and Type 4 handle autodiscovery, multihoming and fast convergence.
B is correct: Type 2 routes are how a PE learns which MAC lives behind which PE, which is exactly what is needed instead of flooding every unknown destination. C is correct: with ingress replication, a Type 3 route per EVI tells the remote PEs to join a P2MP tree for that EVI so broadcast, unknown unicast and multicast are forwarded along it. A and D describe transport choices, and EVPN in an MPLS network runs over plain BGP VPNv4 with no requirement for segment routing or for a VXLAN data plane.
Type 2 = MAC reachability, Type 3 = BUM replication. Anything that is a transport or encapsulation choice (segment routing, VXLAN) is never a prerequisite for EVPN itself.
AC is correct. vrf-target target:65100:100 does two things: it tags routes advertised toward the remote PEs, and on import it selects which of those routes land in this VRF.
vrf-target target:65100:100 does two things: it tags routes advertised toward the remote PEs, and on import it selects which of those routes land in this VRF.
A route target is an extended community in the form target:<admin>:<assigned>. In an vrf-target statement Junos both attaches the community on export to the MP-BGP update sent to remote PEs and matches it on import. That is how the VPN stays private: routes from other VPNs simply do not carry your target, so they are not imported.
C is correct because when the PE exports VPN routes to its remote-PE peers it stamps target:65100:100 on them; without the community the remote site could not tell which routes belong to this VPN. A is the mirror of that statement: on the way in, the same value is used to identify which BGP-learned routes belong to this VRF, so it identifies routes learned from the remote PE. Together they are the two halves of one mechanism, which is why this is a choose-two.
Route targets are an MP-BGP control-plane mechanism. They only ever appear on PE-PE updates, never on PE-CE updates.
Want the full bank of 59 questions for Service Provider Routing and Switching, Professional (JNCIP-SP)? See all practice exams.