Prepare and test your skills

Prepare and test your skills

Free sample question with answers and detailed explanation below. The Professional Cloud Network Engineer practice bank has an additional 2,437 practice questions.Try 10 free realistic questions: instant score, no credit card, and no account required.
You are deploying a private VPC-native GKE Standard cluster in us-central1 where Pods experience highly variable outbound bursts to an external partner API. You must meet the following requirements:
How should you configure the GKE cluster and Cloud NAT gateway?
Keeping default node-level SNAT enabled relies on the ip-masq-agent on each GKE node to masquerade Pod source IP addresses to the host node's primary IP address before traffic enters the VPC. Endpoint-independent mapping (EIM) is a Cloud NAT behavior that consistently maps a specific internal IP and source port to the same external NAT IP and port regardless of the destination address.
ip-masq-agent), and a second time at the Cloud NAT gateway. This hides original Pod IP addresses from VPC Flow Logs and Packet Mirroring for internet-bound flows and keeps unnecessary conntrack state on the nodes.TIME_WAIT tuning for same-destination bursts: Relying on default 120s TCP TIME_WAIT timeouts instead of lowering the timeout to 5s increases the risk of port exhaustion when Pods open rapid sequential connections to the same partner endpoint.ENDPOINT_INDEPENDENT_CONFLICT packet drops when multiple connections contend for reused mappings.Keeping default SNAT would be appropriate only if you needed to hide Pod IP addresses from on-premises networks over Cloud VPN or Cloud Interconnect, and EIM would only be valid if the gateway used static port allocation instead of dynamic port allocation.
--disable-default-snat, configure Public Cloud NAT for all primary and secondary IP ranges of the cluster subnet using manual NAT IP allocation, enable dynamic port allocation (minimum 64, maximum 2048 ports per VM), and reduce the TCP TIME_WAIT timeout to 5s.--disable-default-snat with Cloud NAT accomplishesCombining --disable-default-snat on a VPC-native GKE cluster with a Public Cloud NAT gateway covering both primary and secondary subnet ranges eliminates redundant node-level Source NAT (SNAT) while preserving Pod IP visibility across VPC telemetry and offloading SNAT tracking to Google Cloud's distributed Andromeda networking stack.
ip-masq-agent from translating Pod IP addresses into node IP addresses before packets enter the VPC. Because Pod secondary alias IP addresses remain intact inside the VPC, VPC Flow Logs, Packet Mirroring, and Firewall Rules Logging record the actual originating Pod IP addresses, while Cloud NAT performs a single SNAT translation at egress.TIME_WAIT timeout from 120s to 5s releases closed socket tuples rapidly when many simultaneous connections target the same destination IP, port, and protocol (3-tuple).iptables to Cloud NAT.Unlike keeping default SNAT enabled or using automatic IP allocation, this architecture provides deterministic allowlistable egress IPs, full Pod-level observability inside the VPC, and adaptive per-node port utilization without manual intervention during traffic spikes.
--disable-default-snat, configure Public Cloud NAT for all primary and secondary IP ranges of the cluster subnet using automatic NAT IP allocation, set static port allocation to 64 minimum ports per VM, and reduce the TCP TIME_WAIT timeout to 5s.Automatic NAT IP address allocation delegates external IP provisioning to Google Cloud, dynamically reserving and releasing regional external IPv4 addresses as node counts fluctuate. Static port allocation assigns a fixed, unchanging pool of source ports to every GKE node VM regardless of actual outbound connection demand.
/32), Cloud NAT automatically overrides a configured 64-port minimum and reserves at least 1,024 source ports per node VM. Idle nodes tie up 1,024 ports each, while high-concurrency nodes cannot dynamically scale beyond that fixed limit without manual reconfiguration.OUT_OF_RESOURCES errors.This setup would be suitable only if the external partner did not require IP allowlisting, all GKE nodes had uniform, predictable egress concurrency within 1,024 ports per node, and you did not need to preserve or drain specific static external IPs.
--disable-default-snat, configure Public Cloud NAT for only the primary IP range of the cluster subnet using manual NAT IP allocation, enable dynamic port allocation (minimum 64, maximum 2048 ports per VM), and reduce the TCP TIME_WAIT timeout to 5s.Configuring a Public Cloud NAT gateway to apply solely to the primary IPv4 address range of a subnet restricts SNAT translation to traffic whose source IP belongs to the Compute Engine VM primary interface addresses—meaning GKE node IP addresses—while ignoring alias IP ranges allocated from secondary subnet ranges.
--disable-default-snat is enabled on a VPC-native GKE cluster, Pods retain their secondary IPv4 addresses as the packet source IP when traffic exits the node into the VPC network. Because Cloud NAT is not configured to serve the Pod secondary IP range, it drops or bypasses translation for those packets, preventing Pods from reaching the external partner API.5s TCP TIME_WAIT timeout are properly tuned, no Pod traffic matches the NAT gateway's source subnet filter.ip-masq-agent rewrites Pod source IPs to node primary IPs on every GKE node.This configuration would work only if default SNAT remained enabled (or ip-masq-agent masqueraded internet destinations to node IPs), which would violate the requirement to avoid double SNAT and preserve Pod IP visibility in VPC Flow Logs.
For VPC-native GKE clusters using --disable-default-snat, Cloud NAT must cover at least the Pod secondary IP range (and ideally both the primary and secondary ranges of the cluster subnet) so both node-level system traffic and Pod workload traffic can egress.
Answer them yourself and get your score the moment you finish. No account needed.
Full access includes
Keep the momentum going with these hand-picked practice scenarios
Last updated Validated