Question-34: Compute Engine provides customised virtual machines that are hosted in Google's data centres. These virtual machines have access to high-performance networking infrastructure and block storage options. App Engine is a platform for the development and hosting of scalable web apps that is completely managed and does not need a server. Your operations team has requested your assistance in diagnosing a performance problem that has been seen in a production application that is hosted on Compute Engine. When there is a lot of demand on the application, it will reject some of the requests that come in. A single application process is shown in the process list for affected instances. This application is using up all of the available CPU, and automatic scaling has reached its maximum number of instances. There is no unusual load on any of the other systems that are connected to this one, including the database. You want to get production traffic back up as soon as you possibly can so that it may be serviced. Which course of action would you suggest I take?
A. Make the agent.googleapis.com/memory/percent used measure the auto scaling metric.
B. On a staggered schedule, restart the instances that were impacted by the issue.
C. Use SSH to connect to each instance, then restart the application process.
D. Raise the limit on the most instances that may be included in the autoscaling group.
Correct Answer
Get All 340 Questions and Answer for Google Professional Cloud Architect
: 4 Explanation: Option 1, since changing the metric that is used for autoscaling will not solve the problem, the CPU is already overutilized, and the only unique workaround available meanwhile the application that is causing the problem is fixed (connection leaks, infinite loops, etc.), is to allow the introduction of new nodes, workers, or virtual machines. It is entirely dependent on the method of diagnosis and repair that you choose to implement. Do you want to examine the application before raising the maximum number of instances in the scaling group, or after increasing the maximum number of instances? If this happens in the real world, I imagine that consumers would ask for an increase in the maximum number of instances, and if the application process continues to use up all of the CPU, they will most likely close and reopen the programme. Option 4 is the only one that makes any sense. A is not a valid choice. Two, you could restart, but you have no guarantee that this would solve the problem. C SSH assumes unix vm's (?). The maximum number of instances that can be scaled automatically has been reached. There is no unusual load on any of the other systems that are connected to this one, including the database. The true question is whether or not the application running on the container has reached its maximum capacity and whether or not it is time to generate fresh instances. Option 4 We haven't found anything that mentions utilising memory measurements to determine whether or not a significant percentage of the CPU is being used. This is the notion of comparing apples and oranges. There is no indication that the amount of RAM being used is a problem. However, since autoscaling has reached the top limit of instances, increasing the number of instances in a group would make the issue easier to solve (group maxed out). It's possible that the CPU use wouldn't be very high, but the work would be spread out among multiple instances, which would immediately contribute additional CPU power. Although I do not believe this would cover Google's site reliability engineering, it does make it possible for production traffic to once again be provided while the underlying problem is investigated.