Sample questions
Gcp Compute Gce CloudrunDifficulty 1
In Compute Engine, what does a machine type's series/family (e.g. e2, n2, c2) primarily determine?
- aOnly the region the VM can be created in, not its hardware profile
- bThe underlying CPU platform and the balance of price/performance/workload optimization (general-purpose, compute-optimized, memory-optimized, etc.)✓
- cWhether the VM is allowed to use Persistent Disk at all
- dThe maximum number of VMs allowed per project, regardless of size
Explanation:Machine series group VMs by CPU platform and price/performance profile — e2 is a cost-optimized general-purpose series, n2 a balanced general-purpose series, c2 compute-optimized, and so on. It does not gate region availability by itself (a), does not control Persistent Disk eligibility (c), and has nothing to do with per-project VM quotas (d).
Gcp Compute Gce CloudrunDifficulty 1
What is a custom machine type in Compute Engine?
- aA VM configuration where the operator specifies vCPU count and memory independently instead of picking a predefined size✓
- bA VM type that can only run custom, self-compiled kernels
- cA machine type reserved exclusively for GPU workloads
- dA machine type that automatically changes its vCPU count based on live CPU load
Explanation:Custom machine types let you pick vCPU and memory amounts directly (within allowed ratios) rather than choosing one of Google's predefined shapes, which is useful when a workload's needs don't match a standard size. It has nothing to do with kernel customization (b), is not GPU-exclusive (c), and does not auto-resize live (d) — resizing a running instance's machine type requires a stop/start.
Gcp Compute Gce CloudrunDifficulty 1
What is the main tradeoff of using a Spot VM in Compute Engine compared to a standard on-demand VM?
- aSpot VMs cost more but come with a guaranteed uptime SLA
- bSpot VMs can only run for a maximum of one hour before mandatory deletion
- cSpot VMs are significantly cheaper but Google can preempt the instance with short notice when it needs the capacity back✓
- dSpot VMs cannot attach any Persistent Disk and only support local SSD
Explanation:Spot VMs offer a substantial discount in exchange for no availability guarantee — Google can reclaim the capacity and terminate the instance with short notice whenever needed. There is no SLA (a is backwards), no strict one-hour cap in general (b), and Spot VMs can still attach Persistent Disks (d).
Gcp Compute Gce CloudrunDifficulty 2
In a Managed Instance Group (MIG), what is the purpose of a health check when used with autoscaling/autohealing?
- aIt measures only the VM's boot disk free space and nothing else
- bIt is only used to decide which region to create new instances in
- cIt determines the billing tier applied to each instance in the group
- dIt periodically probes each instance's application-level health✓
Explanation:A MIG health check (distinct from basic VM status) periodically checks whether each instance is actually serving correctly (e.g. via an HTTP probe); if an instance fails the check for a sustained period, autohealing can automatically delete and recreate it. It isn't tied to disk space specifically (a), region placement (b), or billing (c).
Gcp Compute Gce CloudrunDifficulty 2
In Cloud Run, what does the request concurrency setting control?
- aThe total number of Cloud Run services allowed in one project
- bThe maximum number of simultaneous requests a single container instance is allowed to handle at once✓
- cThe number of retries Cloud Run performs before returning an error to the client
- dThe maximum size in megabytes of a single request payload
Explanation:Concurrency defines how many in-flight requests one container instance can process concurrently before Cloud Run routes additional requests to another instance (or spins one up). It does not limit project-wide service count (a), is unrelated to retry behavior (c), and does not set a payload size limit (d).
Gcp Compute Gce CloudrunDifficulty 2
What causes a "cold start" in a Cloud Run or Cloud Functions service?
- aA new container/execution environment must be provisioned and initialized✓
- bThe VM's CPU temperature drops below an operating threshold and must warm up
- cThe service's region has to be changed to match the caller's location
- dThe billing account requires re-authentication before any request can be processed
Explanation:A cold start happens when there is no already-running ("warm") instance ready to serve a request, so the platform must provision a new container, load the application, and run any startup code before the first response — adding latency compared to a warm instance. It is not about literal hardware temperature (b), regional failover (c), or billing re-auth (d).