Question 370
An instance rebooted unexpectedly and administrators need to compare recent platform actions. What is the most direct solution?
- Create a custom Compute Engine image from a prepared source disk
- Inspect the instance's serial port output and boot diagnostics
- Review guest startup-script logs and execution results
- Inspect relevant Cloud Logging audit entries and instance operations
- Collect appropriate guest process telemetry using Ops Agent or OS tools
Explanation
Correct Answer
D:Audit and operation records can reveal who initiated supported management actions. Guest application logs alone may not capture control-plane changes.
Incorrect Answers
A:Choosing this would create a reusable VM boot image, an appropriate action for vm image operations operations; It cannot take the place of the separate step needed to investigate VM control-plane restarts in this specific situation.
B:This approach can inspect early VM boot output. The requirement is to investigate VM control-plane restarts, so it controls a different part of the workflow and leaves that result unaddressed.
C:This choice is useful to troubleshoot startup-script execution. The stated task is to investigate VM control-plane restarts, which requires a different mechanism or permission scope.
E:Using this action would inspect VM guest process consumption, not investigate VM control-plane restarts. Its function is different from the specific change described in the scenario.
Question 371
A Linux Compute Engine VM abruptly becomes inaccessible by SSH after a guest OS change. Its instance status is RUNNING. Which THREE investigations can help isolate the failure without immediately deleting the boot disk? (Choose THREE.)
- Inspect serial console or serial port output for guest boot failures
- Assume the Cloud Billing budget email threshold is the mechanism rejecting each SSH packet
- Delete the boot disk before inspecting any logs to force a clean login state
- Check OS Login and guest SSH configuration for the calling identity
- Check VPC firewall rules and the route/path used for the selected SSH connectivity method
Explanation
Correct Answers
A:Serial output can reveal bootloader, kernel, and startup errors when SSH is unavailable. It is useful even when the guest cannot complete normal network login.
D:Guest login policy, IAM authorization and the SSH service itself can block access. These controls are independent of the Compute Engine instance's RUNNING state.
E:A running VM can remain unreachable because network policy or a required connection path denies SSH traffic. The appropriate path depends on IAP, internal IP, or external access.
Incorrect Answers
B:Billing budget notifications do not enforce guest SSH authorization or VPC firewall decisions. The connection failure must be investigated at guest, IAM, and network layers.
C:Deleting the disk destroys the very guest state and logs needed for diagnosis. Safer troubleshooting reads available boot information and preserves recoverability first.
Question 372
A VM performance issue may involve a particular guest process rather than host-level CPU totals. What should the administrator do?
- Review guest startup-script logs and execution results
- Inspect relevant Cloud Logging audit entries and instance operations
- Run gcloud compute instances describe in its zone
- Collect appropriate guest process telemetry using Ops Agent or OS tools
- Inspect the instance's serial port output and boot diagnostics
Explanation
Correct Answer
D:Guest-level process monitoring gives detail beyond hypervisor CPU metrics. Platform-level charts alone cannot identify every in-guest process bottleneck.
Incorrect Answers
A:This choice is useful to troubleshoot startup-script execution. The stated task is to diagnose process-specific resource use, which requires a different mechanism or permission scope.
B:It can correlate VM management activity with logs; however, it does not provide the specific operation to diagnose process-specific resource use. The two controls serve different purposes.
C:An operator would use this action to retrieve full VM resource configuration, chiefly when trying to inspect one instance's properties; The current requirement is to diagnose process-specific resource use, which remains unresolved by that operation.
E:Using this action would inspect early VM boot output, not diagnose process-specific resource use. Its function is different from the specific change described in the scenario.
Question 373
A support engineer needs to see all GKE clusters in a project and their basic configuration. Which solution should be used?
- Run kubectl get services in the relevant namespace
- Run kubectl get pods with suitable namespace and output options
- List GKE clusters with gcloud container clusters list
- Run kubectl get nodes using the intended cluster context
Explanation
Correct Answer
C:Cluster inventory reports cluster locations, modes and states visible to the caller. It does not list each application's Pods until Kubernetes access is used.
Incorrect Answers
A:Its purpose is to list Kubernetes Service endpoints. That distinction matters: the scenario calls for an action to discover GKE environments in a project, not this separate configuration task.
B:Administrators use it to list Kubernetes Pods and states, rather than to discover GKE environments in a project. Applying it would change the wrong component of this deployment.
D:This approach can inspect cluster worker node readiness. The requirement is to discover GKE environments in a project, so it controls a different part of the workflow and leaves that result unaddressed.
Question 374
An engineer needs the names and readiness status of Kubernetes worker nodes. How should the team proceed?
- Run kubectl get services in the relevant namespace
- Run kubectl get nodes using the intended cluster context
- List GKE clusters with gcloud container clusters list
- Run kubectl get pods with suitable namespace and output options
- Run gcloud compute instances list with the correct project
Explanation
Correct Answer
B:The Kubernetes API reports node objects and readiness. A Compute Engine instance list is not a universal substitute for GKE node state.
Incorrect Answers
A:It can list Kubernetes Service endpoints; however, it does not provide the specific operation to check Kubernetes nodes. The two controls serve different purposes.
C:Select this when the goal is to list managed Kubernetes clusters. Here the decisive requirement is to check Kubernetes nodes, and that outcome is not achieved by this action.
D:The tool can list Kubernetes Pods and states, but it cannot be substituted for the capability needed to check Kubernetes nodes. Its operational effect is not equivalent.
E:This proposal can list VM inventory within a project, but the failure or goal described here calls for a way to check Kubernetes nodes; It changes the wrong control or scope for the requested outcome.
Question 375
A new release is deployed but the operator wants to list its current Pods and restart counts. What is the most direct solution?
- List GKE clusters with gcloud container clusters list
- Run kubectl get services in the relevant namespace
- Run kubectl get nodes using the intended cluster context
- Run kubectl get pods with suitable namespace and output options
- Use gcloud compute ssh over a direct reachable address with appropriate access
Explanation
Correct Answer
D:Pod inventory reports workload state and can show restarts. The correct namespace and context are important to avoid missing resources.
Incorrect Answers
A:Using this action would list managed Kubernetes clusters, not inspect deployment Pod health. Its function is different from the specific change described in the scenario.
B:This approach can list Kubernetes Service endpoints. The requirement is to inspect deployment Pod health, so it controls a different part of the workflow and leaves that result unaddressed.
C:This setting would inspect cluster worker node readiness. That behavior does not establish the requested ability to inspect deployment Pod health; it addresses an adjacent concern instead.
E:Its purpose is to start a direct CLI SSH session to a VM. That distinction matters: the scenario calls for an action to inspect deployment Pod health, not this separate configuration task.
Question 376
A platform group needs to see Kubernetes Services and their cluster or external addresses. Which implementation matches this requirement?
- Run kubectl get pods with suitable namespace and output options
- Run kubectl get nodes using the intended cluster context
- Run kubectl get services in the relevant namespace
- List GKE clusters with gcloud container clusters list
- Use the supported GKE node pool upgrade process
Explanation
Correct Answer
C:Services represent stable network access for Pod endpoints. Listing them reveals types and assigned addresses, not the health of every backend by itself.
Incorrect Answers
A:The result of this choice would be to list Kubernetes Pods and states. The requested outcome, to inspect service exposure in a namespace, is handled at a different layer or through another operation.
B:Select this when the goal is to inspect cluster worker node readiness. Here the decisive requirement is to inspect service exposure in a namespace, and that outcome is not achieved by this action.
D:This mechanism is intended to list managed Kubernetes clusters. It does not implement the required action to inspect service exposure in a namespace, even though both are relevant to the same general service.
E:For the stated need to inspect service exposure in a namespace, this would instead upgrade an existing node pool; it supports the separate task to maintain worker node versions, not the required configuration.
Question 377
A container repeatedly exits after startup and its Pod is in CrashLoopBackOff. Which action is most appropriate?
- Check Service selectors against Pod labels and readiness
- Inspect scheduler events and node resource requests or capacity
- Read the Pod's current and previous container logs and describe its status
- Update the Deployment image reference to an existing valid tag or immutable digest
Explanation
Correct Answer
C:Logs can reveal application exceptions; events show restart and probe behavior. Increasing replicas does not fix a crashing container image.
Incorrect Answers
A:It can match Service selectors with ready Pods; however, it does not provide the specific operation to investigate CrashLoopBackOff. The two controls serve different purposes.
B:This setting would diagnose scheduling resource insufficiency. That behavior does not establish the requested ability to investigate CrashLoopBackOff; it addresses an adjacent concern instead.
D:Using this action would correct a nonexistent container image reference, not investigate CrashLoopBackOff. Its function is different from the specific change described in the scenario.
Question 378
A Pod is stuck in Pending because no node meets its CPU request. Which solution should be used?
- Check Service selectors against Pod labels and readiness
- Read the Pod's current and previous container logs and describe its status
- Update the Deployment image reference to an existing valid tag or immutable digest
- Inspect scheduler events and node resource requests or capacity
- Grant an appropriate Artifact Registry writer role to the build identity
Explanation
Correct Answer
D:Scheduling considers resource requests and placement constraints. More container image pulls cannot solve unavailable CPU allocation.
Incorrect Answers
A:The result of this choice would be to match Service selectors with ready Pods. The requested outcome, to resolve a Pending Pod due to capacity, is handled at a different layer or through another operation.
B:While this can diagnose repeated Kubernetes container crashes, the present task requires an action to resolve a Pending Pod due to capacity. This operation does not supply the behavior described in the requirements.
C:Its purpose is to correct a nonexistent container image reference. That distinction matters: the scenario calls for an action to resolve a Pending Pod due to capacity, not this separate configuration task.
E:An operator would use this action to authorize CI image upload, chiefly when trying to fix denied container image push; The current requirement is to resolve a Pending Pod due to capacity, which remains unresolved by that operation.
Question 379
A GKE Deployment references an Artifact Registry image tag that was deleted, and new Pods enter ImagePullBackOff. How should the team proceed?
- Adjust GPU node capacity or Pod GPU resource requests
- Inspect scheduler events and node resource requests or capacity
- Check Service selectors against Pod labels and readiness
- Update the Deployment image reference to an existing valid tag or immutable digest
- Read the Pod's current and previous container logs and describe its status
Explanation
Correct Answer
D:The image reference must identify an artifact version that the registry can return. Adding reader permissions cannot resolve an image tag that no longer exists in the repository.
Incorrect Answers
A:For the stated need to recover Pods from a deleted registry image tag, this would instead align Pod GPU requests with node capacity; it supports the separate task to resolve insufficient accelerator scheduling, not the required configuration.
B:Select this when the goal is to diagnose scheduling resource insufficiency. Here the decisive requirement is to recover Pods from a deleted registry image tag, and that outcome is not achieved by this action.
C:This choice is useful to match Service selectors with ready Pods. The stated task is to recover Pods from a deleted registry image tag, which requires a different mechanism or permission scope.
E:The tool can diagnose repeated Kubernetes container crashes, but it cannot be substituted for the capability needed to recover Pods from a deleted registry image tag. Its operational effect is not equivalent.
Question 380
A workload receives no traffic because its Kubernetes Service has no selected endpoints. What is the most direct solution?
- Update the Deployment image reference to an existing valid tag or immutable digest
- Check Service selectors against Pod labels and readiness
- Read the Pod's current and previous container logs and describe its status
- Inspect scheduler events and node resource requests or capacity
- Run kubectl get services in the relevant namespace
Explanation
Correct Answer
B:A Service selects Pods through labels, with readiness affecting endpoint availability. A healthy load balancer cannot route to absent service endpoints.
Incorrect Answers
A:This approach can correct a nonexistent container image reference. The requirement is to restore Kubernetes backend endpoints, so it controls a different part of the workflow and leaves that result unaddressed.
C:Its purpose is to diagnose repeated Kubernetes container crashes. That distinction matters: the scenario calls for an action to restore Kubernetes backend endpoints, not this separate configuration task.
D:While this can diagnose scheduling resource insufficiency, the present task requires an action to restore Kubernetes backend endpoints. This operation does not supply the behavior described in the requirements.
E:In a workflow for gke cluster inventory operations, it can list Kubernetes Service endpoints; This question instead asks how to restore Kubernetes backend endpoints; substituting the former action would leave that distinct technical requirement unmet.