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.
Your organization runs an application behind a global external Application Load Balancer using regional Managed Instance Groups (MIGs) across three regions. Another backend service shares these MIGs with a 90-second connection draining timeout.
For zero-downtime updates and regional fault isolation, you must:
240 seconds to finish before termination.25% of VMs pass HTTP health checks, if under 50% of backends are degraded.35% or more VMs stay healthy for 60 seconds.Which configuration should you implement?
240-second connection draining with autoCapacityDrain isCombining a 240-second connection draining timeout on the global backend service with a service load balancing policy (serviceLbPolicy) that enables auto-capacity draining coordinates graceful instance-level termination with automated region-level health isolation.
240-second termination: When multiple backend services share the same MIG, Google Cloud enforces the largest connection draining timeout (max(90, 240) = 240 seconds). During rolling updates or scale-in events, existing connections continue on draining VMs while new connections are withheld, delaying MIG VM deletion for at least 240 seconds.<25% health, <50% cap): With auto-capacity draining enabled, any MIG with fewer than 25% of instances passing the custom HTTP health check becomes a drainable candidate backend and has its effective capacity set to 0—without modifying capacityScaler—as long as fewer than 50% of all backend groups are drainable candidates.>=35% for 60 seconds): Auto-capacity undraining restores the backend's capacity only after 35% or more of its instances pass health checks continuously for at least 60 seconds, preventing rapid drain/undrain oscillation.90-second timeout on the sibling backend service.50% maximum drainable-backend ratio.capacityScaler edits during unplanned failures or rolling software deployments.| Mechanism | Trigger | Threshold / Behavior |
|---|---|---|
| Connection Draining | VM stop/delete/removal | Delays VM deletion by 240s (max(90, 240)) |
| Auto-Capacity Drain | Custom HTTP health check | Capacity 0 if <25% healthy and <50% backends degraded |
| Auto-Capacity Undrain | Sustained recovery | Restores capacity at >=35% healthy for 60s |
Together, these native backend service controls satisfy every SLO and zero-downtime requirement without extra proxy infrastructure.
0-second timeout and failoverHealthThreshold doSetting connectionDraining.drainingTimeoutSec to 0 disables connection draining on that backend service, allowing existing connections to end as quickly as possible. The failoverHealthThreshold (configured from 1 to 99, defaulting to 70) in a service load balancing policy dictates when a proxy shifts traffic from primary backends to designated failover backends.
90-second effective timeout: Although the shared backend service keeps the MIG's effective connection draining timeout at 90 seconds (max(0, 90)), it still falls short of the required 240 seconds for active requests during VM replacement.failoverHealthThreshold to 25 only triggers traffic spilling to designated failover backends when healthy endpoints drop below 25%. It does not zero out the backend's effective capacity via auto-capacity draining, enforce the <50% drain ratio limit, or require 35% health for 60 seconds to restore capacity.0-second timeout disables connection draining rather than inheriting a 240-second threshold.failoverHealthThreshold operates on primary-to-failover backend transitions, whereas auto-capacity draining explicitly evaluates drainable candidate backends against the 25% drain, 50% cap, and 35% / 60-second undrain rules.failoverHealthThreshold too low causes the load balancer to keep routing traffic to severely degraded primary backends.In-flight requests finished within 90 seconds and you wanted primary-to-failover backend spilling at 25% health rather than auto-capacity draining and undraining.
150-second connection draining and manual capacityScaler doConnection draining timeouts on a backend service define the minimum duration existing connections can persist before instances are removed or terminated. Meanwhile, capacityScaler is a backend service property (from 0.0 to 1.0) that manually scales a backend's target serving capacity without deleting its instances.
150 seconds instead of 240 seconds): Connection draining timeouts across multiple backend services sharing the same MIG are not cumulative. Google Cloud applies the largest configured timeout value among the shared backend services. Setting 150 seconds alongside a 90-second service results in an effective timeout of 150 seconds (max(90, 150)), cutting off long-running 240-second connections early.capacityScaler requires operator or pipeline intervention before planned updates and provides no automatic reaction when an unplanned fault causes fewer than 25% of instances to pass HTTP health checks.capacityScaler to 0 does not enforce the <50% degraded backend cap, nor does it automatically undrain capacity after 35% of instances pass health checks for 60 seconds.max(timeout_1, timeout_2, ..., timeout_N).capacityScaler overrides persist until explicitly reverted, increasing operational toil during recovery.The maximum in-flight request duration were 150 seconds or less and the team only needed manual, pre-scheduled traffic shifting rather than automated health-check-driven draining and undraining.
Deploying backup regional external Application Load Balancers with a Cloud DNS FAILOVER routing policy establishes global-to-regional infrastructure failover when the global load balancer itself experiences an outage. The --backup-data-trickle-ratio flag continuously routes a small fraction of client traffic to backup regional VIPs to keep them warm.
RATE balancing mode on shared MIGs: When global and regional external Application Load Balancers share the same MIG backends, both backend services must use UTILIZATION balancing mode. Using RATE balancing mode can cause backend VMs to receive 2x traffic simultaneously from both load balancers during failover, risking severe overload.0.35): A 0.35 DNS trickle ratio sends 35% of all global client traffic to backup regional load balancers during steady state; it has nothing to do with waiting for 35% of instances to stay healthy for 60 seconds before restoring capacity.<25% health with a <50% safeguard inside the global backend service.240-second connection draining works as expected, but RATE mode and CPU/autoscaler misconfigurations cause secondary outages during failover.DNS TTL + Health Check Interval * Unhealthy Threshold) make DNS-based switching far slower than internal Envoy/GFE capacity draining.You needed resilience against a global load balancer control-plane outage using UTILIZATION balancing mode, CPU-based autoscaling, and scale-in controls rather than backend-level auto-capacity draining.
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