Skip to content

Professional Cloud Network Engineer · Domain 3: Configuring managed network services · 16% of the exam

Task 3.3: Configuring Cloud DNS.

Zones and records, migrating to Cloud DNS, routing policies, DNSSEC, forwarding and server policies, private and public zones with split horizon, cross-project binding and DNS peering, and DNS for GKE.

Study it

  • Cloud DNS: zones, records, routing policies, DNSSEC, forwarding, server policies, peering and DNS for GKE

    Lesson coming

Sample questions

Try each one before opening the answer. Every option is explained, with the Google Cloud documentation page that proves it.

Question 1 · choose 1

In a Cloud DNS private zone, api.corp.internal has a geolocation routing policy: europe-west1 returns the address of an internal Application Load Balancer in Europe, and us-east1 returns one in the US. Both items are health checked, both load balancers have global access, and geofencing is enabled on the policy. Every backend of the European load balancer fails its health checks. What do VMs in europe-west1 receive when they resolve api.corp.internal?

  1. AThe US load balancer's address, the next closest geolocation
  2. BAn NXDOMAIN response until the European backends pass their health checks
  3. CBoth load balancers' addresses, spread by weighted round robin
  4. DThe European load balancer's address, returned unaltered
Show the answer and why
  • AThe US load balancer's address, the next closest geolocation

    Incorrect

    Failing over to the next closest geolocation is the behavior without geofencing; geofencing turns that failover off.

  • BAn NXDOMAIN response until the European backends pass their health checks

    Incorrect

    As an authoritative server, Cloud DNS must return a value; it doesn't answer NXDOMAIN for an unhealthy item.

  • CBoth load balancers' addresses, spread by weighted round robin

    Incorrect

    A geolocation policy doesn't merge items into a weighted round robin answer when one region fails.

  • DThe European load balancer's address, returned unaltered

    Correct

    With geofencing, traffic stays in its geolocation; when its endpoints fail, Cloud DNS returns their addresses unaltered.

Without geofencing, a geolocation policy fails over to the next closest geolocation when an item's endpoints fail health checks. A geofence keeps traffic in its region even if all endpoints there fail; because Cloud DNS must still answer, it returns those addresses unaltered.

Question 2 · choose 1

A VPC network has IPv6-only subnets and dual-stack subnets in us-east4, and none of the VMs has an external IP address. A DNS64 server policy is bound to the network, and a Public NAT gateway in us-east4 is configured only for the IPv6-only subnets, with NAT64 enabled for them. All VMs must call a partner API whose name has only an A record. The IPv6-only VMs reach it; the dual-stack VMs don't. What should the network engineer change?

  1. AEnable NAT64 on the gateway for the dual-stack subnets' IPv6 ranges as well
  2. BConfigure the gateway to provide NAT for the dual-stack subnets' IPv4 ranges too
  3. CBind the DNS64 server policy to the dual-stack subnets as well
  4. DGive the dual-stack subnets external IPv6 ranges so that their VMs reach the API over IPv6
Show the answer and why
  • AEnable NAT64 on the gateway for the dual-stack subnets' IPv6 ranges as well

    Incorrect

    NAT64 serves IPv6-only interfaces, and the dual-stack VMs receive the A record and send IPv4, so no NAT64 translation is involved.

  • BConfigure the gateway to provide NAT for the dual-stack subnets' IPv4 ranges too

    Correct

    DNS64 doesn't apply to dual-stack VMs, so they connect to the IPv4 address, and without external IP addresses they need Public NAT for their IPv4 ranges.

  • CBind the DNS64 server policy to the dual-stack subnets as well

    Incorrect

    A server policy binds to VPC networks, and DNS64 server policies don't apply to dual-stack VMs at all.

  • DGive the dual-stack subnets external IPv6 ranges so that their VMs reach the API over IPv6

    Incorrect

    The partner API has only an IPv4 address, so external IPv6 addresses give the VMs no path to it.

With a DNS64 server policy, Cloud DNS synthesizes 64:ff9b::/96 addresses for names without AAAA records, and a Cloud NAT gateway with NAT64 translates traffic from IPv6-only interfaces to those addresses. DNS64 server policies don't apply to dual-stack or IPv4-only VMs: they get the A record and connect over IPv4. A VM without an external IP address reaches an IPv4 destination through Public NAT only if the gateway applies to the range its primary internal address comes from.

Question 3 · choose 1

In one organization, an application team wants to own a Cloud DNS private zone in its service project while making it visible to the Shared VPC network of the host project, using cross-project binding. Creating the zone with the host network's URL fails with a permission error, although the team can create zones in its own project. Which permission is missing, and where?

  1. Adns.managedZones.create on the host project
  2. Bcompute.networkUser on the host project
  3. Cdns.networks.bindPrivateDNSZone on the service project
  4. Ddns.networks.bindPrivateDNSZone on the host project
Show the answer and why
  • Adns.managedZones.create on the host project

    Incorrect

    The zone is created in the service project, where the team already holds this permission.

  • Bcompute.networkUser on the host project

    Incorrect

    Binding a zone to a network needs a Cloud DNS permission, not the right to use the network's subnets.

  • Cdns.networks.bindPrivateDNSZone on the service project

    Incorrect

    The bind permission is checked on the project that owns the VPC network, not on the project that owns the zone.

  • Ddns.networks.bindPrivateDNSZone on the host project

    Correct

    Cross-project binding needs dns.networks.bindPrivateDNSZone on the project that owns the VPC network, here the host project.

Cross-project binding lets a managed private zone in one project be bound to a VPC network owned by another project in the same organization. It needs dns.managedZones.create on the project that owns the zone and dns.networks.bindPrivateDNSZone on the project that owns the network; the DNS Administrator role includes both.

Question 4 · choose 1

A VPC network's subnets use privately used public IPv4 ranges. Reverse lookups (PTR queries) for the VMs' addresses fail, because Cloud DNS sends them to the internet. What should the network engineer create?

  1. AA public zone for the matching in-addr.arpa name, with PTR records added by hand
  2. BA forwarding zone for the in-addr.arpa name that targets the internet resolvers
  3. CA response policy rule that answers the PTR queries with fixed data
  4. DA managed reverse lookup zone whose PTR answers come from Compute Engine data
Show the answer and why
  • AA public zone for the matching in-addr.arpa name, with PTR records added by hand

    Incorrect

    A public zone publishes records to the internet; the VMs' addresses must be resolved privately from Compute Engine data.

  • BA forwarding zone for the in-addr.arpa name that targets the internet resolvers

    Incorrect

    Sending the queries out is what fails today; the answers must come from Compute Engine's own data.

  • CA response policy rule that answers the PTR queries with fixed data

    Incorrect

    Fixed answers would not follow the VMs' addresses as they change.

  • DA managed reverse lookup zone whose PTR answers come from Compute Engine data

    Correct

    Managed reverse lookup zones let Cloud DNS resolve non-RFC 1918 PTR records for VMs locally instead of over the internet.

By default Cloud DNS forwards PTR queries for non-RFC 1918 addresses through the public internet. A managed reverse lookup zone is a private zone with a special attribute that tells Cloud DNS to answer PTR queries from Compute Engine DNS data; records can't be added to it directly.

Question 5 · choose 1

Customers on the internet resolve app.example.com to the public address of a load balancer. VMs in the company's VPC network must resolve the same name to an internal address instead. How should Cloud DNS be set up?

  1. AA public zone and a private zone for example.com, each with its own app record
  2. BOne public zone with two app records, and Cloud DNS picks the internal one for VMs
  3. CA private zone only, with the public address added as a second app record
  4. DA public zone plus a DNS peering zone that sends VM queries to another network
Show the answer and why
  • AA public zone and a private zone for example.com, each with its own app record

    Correct

    This split horizon setup answers queries from the authorized network from the private zone and everyone else from the public zone.

  • BOne public zone with two app records, and Cloud DNS picks the internal one for VMs

    Incorrect

    A public zone gives every client the same answers; it can't tell VMs apart from internet clients.

  • CA private zone only, with the public address added as a second app record

    Incorrect

    A private zone is visible only to authorized networks, so internet customers couldn't resolve the name.

  • DA public zone plus a DNS peering zone that sends VM queries to another network

    Incorrect

    Peering delegates resolution to another network; it doesn't by itself give the internal answer.

Split horizon DNS combines a public zone and a private zone for the same domain. VMs in the networks authorized for the private zone get its records; queries from anywhere else are answered from the public zone.

Google Cloud documentation

Question 6 · choose 1

A GKE cluster uses Cloud DNS for GKE with cluster scope. Compute Engine VMs in the same VPC network and on-premises clients reached over Cloud VPN now need to resolve the cluster's headless Services, while Pods keep cluster-scoped resolution. What should the network engineer enable?

  1. ANodeLocal DNSCache on every node pool
  2. BA switch from Cloud DNS for GKE back to kube-dns
  3. CAn outbound server policy that points VMs at the cluster's nodes
  4. DThe Cloud DNS additive VPC scope for the cluster
Show the answer and why
  • ANodeLocal DNSCache on every node pool

    Incorrect

    NodeLocal DNSCache speeds up lookups from Pods; it doesn't publish Service names to VMs or on-premises clients.

  • BA switch from Cloud DNS for GKE back to kube-dns

    Incorrect

    kube-dns serves Pods inside the cluster; it doesn't make Service names resolvable from other VPC resources.

  • CAn outbound server policy that points VMs at the cluster's nodes

    Incorrect

    Alternative name servers in an outbound policy aren't how GKE publishes Service records to the VPC network.

  • DThe Cloud DNS additive VPC scope for the cluster

    Correct

    Additive VPC scope extends cluster scope to make headless Services resolvable from VMs and from on-premises clients.

Cloud DNS for GKE offers cluster scope, VPC scope and the optional additive VPC scope. Additive VPC scope is enabled alongside cluster scope and makes headless Services resolvable from other resources in the VPC network, such as Compute Engine VMs or on-premises clients connected through Cloud VPN or Cloud Interconnect.

Practice domain 3 →Practice all domains →