Google Professional Cloud DevOps Engineer Question 173
Select 3Google Cloud PlatformYour team is running a large-scale batch processing workload on Google Cloud. To minimize costs, you decide to use Spot VMs. However, during a critical phase of processing, you notice that the workload is interrupted because some of the Spot VMs were preempted. How can you ensure the workload continues processing while still optimizing costs?
- A
Use managed instance groups (MIGs) with Spot VMs configured and enable automatic scaling to replace preempted instances.
- B
Configure a backup pool of regular (non-Spot) VMs in a managed instance group to handle critical phases of the workload.
- C
Increase the priority of the Spot VMs to reduce the likelihood of preemption.
- D
Implement checkpointing in your batch processing application to allow it to resume from the last saved state after a preemption.
- E
Manually monitor the Spot VM preemptions and restart the workload on new Spot VMs as needed.
Show answer and explanation
Correct answers: A, B, D
Explanation
Spot VMs are cost-effective but are subject to preemption based on resource availability. To optimize for reliability and cost, you can use managed instance groups (MIGs) to automatically replace preempted Spot VMs, configure a backup pool of regular VMs for critical phases, and implement checkpointing in your application to handle interruptions gracefully. These strategies allow the workload to continue processing with minimal disruption and optimized costs.
- A. Correct.
Using managed instance groups (MIGs) with Spot VMs and enabling automatic scaling allows preempted instances to be automatically replaced, ensuring minimal downtime for your workload.
- B. Correct.
Configuring a backup pool of regular VMs provides a fallback to handle critical phases, ensuring that the workload continues even if all Spot VMs are preempted.
- C. Incorrect.
Spot VMs do not have a priority setting to reduce preemption likelihood. Preemption is determined by Google Cloud's capacity availability, not by instance priority.
- D. Correct.
Implementing checkpointing in the application ensures that the workload can resume from the last saved state, avoiding complete reprocessing after a preemption.
- E. Incorrect.
Manually monitoring and restarting workloads is time-consuming and error-prone, making it unsuitable for large-scale batch processing workloads.