• Option A is incorrect. Cloud Profiler is used to continuously gather information such as CPU and memory usage from applications.
• Option B is CORRECT. Cloud Trace helps you to understand how long your application takes to process requests and the overall latency of requests.
• Option C is incorrect. Cloud Logging doesn’t show the latency or bottlenecks in your applications. It simply shows application logs as generated by the code.
• Option D is incorrect. Cloud Debugger allows you to inspect the application code’s state while it is running .

Reference

https://cloud.google.com/trace/docs/overview

Without the logging agent VMs cannot write logs to Cloud Logging.

• Option B is incorrect. If a logging agent is installed, The VM can send logs to Cloud Logging.
• Option C is incorrect. The default service account has the editor role attached. It has sufficient permissions to write to Cloud Logging.
• Option D is incorrect. The default service account has the editor role attached. It has sufficient permissions to write to Cloud Logging.

Reference

https://cloud.google.com/logging/docs/agent/logging

Cloud Profiler provides insight into how CPU and memory is consumed by applications. This will help with understanding why your cluster is scaling out.

• Option B is incorrect. Cloud Trace shows the latency data and how requests flow through your application.
• Option C is incorrect. Cloud logging is used to collect logging data from applications and services, it does not show the root cause of the scaling action in your cluster.
• Option D is incorrect. Cloud Monitoring provides a centralised view of metrics that can be monitored for GCP services. It does not show the root cause of the scaling action in your cluster.

Reference

https://cloud.google.com/profiler/docs/about-profiler

https://cloud.google.com/monitoring

Committed use instances give the lowest cost, it has a commitment of 1 or 3years. The project will run for 12 months which can make use of the discount from committed use.

• Option B is incorrect. Preemptible instances provide the lowest cost and are excellent for batch processing. But this question specifies the transactions cannot handle interruptions.
• Option C is incorrect. The standard instance is the most expensive option.
• Option D is incorrect. This does not optimize the cost because of the overhead of re-platforming the application

Reference

https://cloud.google.com/compute/docs/instances/preemptible

https://cloud.google.com/compute/docs/instances/signing-up-committed-use-discounts

• Option A is incorrect. Firewall logs can be disabled.
• Option B is incorrect. VPC Flow Logs can be disabled.
• Option C is CORRECT. Policy Denied logs cannot be disabled.
• Option D is incorrect. Data Access logs can be disabled.

Reference

https://cloud.google.com/logging/docs/audit#types

SREs (Site Reliability Engineers) of Google Cloud Platform should spend 50% of their time on engineering work, such as improving the reliability and scalability of the systems they are responsible for, and the other 50% of their time on operational tasks. Therefore, the correct answer is "50%".

"75%" - SREs of Google Cloud Platform do not spend 75% of their time doing operations, as this would not leave enough time for them to focus on engineering work and long-term improvements to the systems they manage.

"SREs of Google Cloud Platform spend 0% of their time doing operations" - SREs of Google Cloud Platform do not spend 0% of their time doing operations, as this would not allow them to respond to operational issues and incidents.

"25%" - SREs of Google Cloud Platform do not spend 25% of their time doing operations

The exam contains some basic recall around logging, this is one of the facts that you must know :)

The correct answer is "There is no logging agent installed on the instance" - Without the logging agent VMs cannot write logs to Cloud Logging. The logging agent is a software package that allows the instance to write to Cloud Logging. If a logging agent is installed, The VM can send logs to Cloud Logging. The default service account has the editor role attached. It has sufficient permissions to write to Cloud Logging. Therefore, the service account does not need the 'logging viewer' role, and the scopes of this service account are sufficient to write to Cloud Logging.

References: https://cloud.google.com/logging/docs/agent/logging

"VM machine logs have not been turned on" - The logging agent is responsible for sending logs from the VM to Cloud Logging. Turning on the VM machine logs does not affect the ability of the logging agent to write to Cloud Logging.

"The service account does not have the 'logging viewer' role" - The logging viewer role is not necessary for writing to Cloud Logging. The default service account has the editor role attached. It has sufficient permissions to write to Cloud Logging.

"The scopes of this service account are insufficient to write to Cloud Logging" - The default service account has the editor role attached. It has sufficient permissions to write to Cloud Logging.

The question asks for the best way to properly shut down an application before the virtual machine instances are pre-empted. Pre-emption occurs when Google reclaims the resources of a virtual machine instance, so it is important to have a way to gracefully shut down an application before this happens.

The correct answer is: "Create a shutdown script and use it as the value for a new metadata entry with the key 'shutdown-script' in the Cloud Platform Console when you create the new virtual machine instance" - This is the best way to ensure that the application can be shut down gracefully before the instance is pre-empted. A shutdown script can be created and used as the value for a new metadata entry in the Cloud Platform Console, allowing the application to be shut down properly when the instance is pre-empted.

References: https://cloud.google.com/compute/docs/shutdownscript

"Create a shutdown script and use it as the value for a new metadata entry with the key 'shutdown-script' in the deployment manager" - While the deployment manager can be used to deploy and manage the virtual machine instances, it cannot be used to set a shutdown script as a metadata entry. The Cloud Platform Console must be used for this purpose.

"Use the CLI, Console or API to send your shut-down instructions before the instances are pre-empted" - While the CLI, Console or API can be used to send shut-down instructions, this is not the best way to ensure that the application can be shut down gracefully before the instance is pre-empted and introduces a huge amount of toil.

"Use the CLI, Console or API to move your workload before the instance is pre-empted" - This is not the best way to ensure that the application can be shut down gracefully before the instance is pre-empted. This introduces toil; additionally, if the aim was to migrate to a new host in event of pre-emption, Google provides pre-emption policies that allow you to do that.

The correct answer is "Discover what users expect from the system and choose SLIs to measure it."

This approach aligns with Google's recommended best practice for selecting SLIs because it involves understanding the specific expectations and needs of the users of the system and using this understanding to guide the selection of appropriate SLIs. By focusing on what users expect from the system, the team can ensure that they are choosing SLIs that are relevant and meaningful to the users, and that will allow them to effectively monitor the system's performance in a way that is meaningful to the users.

"Choose as many SLIs as possible to cover all aspects of the system that users interact with" - This answer is incorrect because choosing too many SLIs can be counter-productive. Having too many SLIs can make it difficult for the team to pay the right level of attention to the indicators that are most important, and can result in an overwhelming amount of data that is difficult to interpret and act upon.

"Choose very few SLIs as possible to reduce unactionable alerts" - This answer is incorrect because choosing too few SLIs can leave significant behaviours of the system unexamined. In order to effectively monitor the system and ensure that it is meeting the required SLOs, it is important to have a sufficient number of SLIs that cover a broad range of aspects of the system's performance.

"Allow users to feedback on issues before choosing SLIs" - This answer is incorrect because it is not a recommended best practice for selecting SLIs. While user feedback can be an important source of information for understanding the needs and expectations of the users, it is not the primary factor that should be considered when selecting SLIs. Instead, the team should focus on understanding what users expect from the system and choose SLIs that are relevant and meaningful to the users.