Mid questions

You are setting up a GitOps pipeline. repo-a contains the application source code, and repo-b contains the Kubernetes configuration (values.yaml).

When code is pushed to repo-a, the pipeline must:

1. Build a Docker image tagged with the Short Git SHA.
2. Automatically update repo-b to use this new image tag.

A repository at /home/interview/deploy-repo contains a values.yaml file tracking the current Docker image tag and a deploy.sh script that simulates production deployment. When deployments fail, the team must manually revert values.yaml to the previous working tag. An automated rollback mechanism is needed to detect failures and restore the last known-good tag with a commit by CI Bot. A starter workflow file has been created at .github/workflows/deploy.yml with the basic structure.

A repository at /home/interview/repo contains a Dockerfile but no automated build process. Developers manually build Docker images without consistent tagging, making it difficult to track which image corresponds to which code version. A starter workflow file has been created at .github/workflows/build.yml with the basic structure.

A repository at /home/interview/repo contains a Node.js application with a test suite at ./run-tests.sh. The application needs to be tested across multiple Node.js versions (18, 20, and 22) to ensure compatibility, but currently has no automated testing configured. A starter workflow file has been created at .github/workflows/pr-tests.yml with the basic structure. The repository has the upload-artifact action available locally at .github/actions/upload-artifact for artifact management.

A repository at /home/interview/repo contains a workflow file .github/workflows/pipeline.yml. The workflow currently defines three jobs: lint, test, and build. These jobs are configured to run in parallel without any dependencies.

A repository at /home/interview/repo contains a Node.js application with a test suite at ./run-tests.sh. The development team needs a two-stage CI/CD pipeline where the first job runs tests and generates test results, and a second job downloads those results to create a summary report. The repository has upload-artifact and download-artifact actions available locally at .github/actions/ for artifact management. A starter workflow file has been created at .github/workflows/artifact-handoff.yml with the basic structure.

A repository at /home/interview/repo contains a multi-component application with infrastructure code in the /infra directory and documentation in the /docs directory. Currently, the repository has no automated workflow configured with path-based filtering to run workflows only when specific files change. A starter workflow file has been created at .github/workflows/infra-check.yml with the basic structure. The repository has the upload-artifact action available locally at .github/actions/upload-artifact for artifact management.

A repository at /home/interview/repo has a test suite located at ./run-tests.sh but no automated testing on pull requests. Developers must manually run tests before merging, leading to inconsistent test coverage and occasional bugs in the main branch. A starter workflow file has been created at .github/workflows/pr-tests.yml with the basic structure. The repository already has the upload-artifact action available locally at .github/actions/upload-artifact for artifact management.

A repository at /home/interview/repo contains multiple applications that share common build and test steps. The development team wants to eliminate code duplication across workflows by creating a centralized reusable workflow that can be called from different workflows with different parameters.

We need to prepare infrastructure for ECS tasks and EC2 instances. It has to span across at least two Availability Zones. These workloads require outbound internet access to download updates and call external APIs. However, inbound access is not allowed. Additionally application should send data to S3 in cost effective way and so we have to deploy necessary infrastructure for that traffic too.

Note: You are required to design the VPC networking architecture only. Creation of ECS clusters, services, or EC2 instances is not part of this task.

You have a container image myapp:grace built from /home/interview/Dockerfile that runs an application requiring cleanup on shutdown. When you run docker stop, the container is killed after the 10-second timeout instead of shutting down gracefully because the application doesn't properly handle the SIGTERM signal.

A Dockerfile at /home/interview/Dockerfile builds a Python web application tagged as myapp:secure. The container runs as the root user (UID 0) with extensive Linux capabilities, violating the principle of least privilege.

A container named mem_test running from image myapp:mem contains a memory stress script at /app/stress_memory.sh. Currently, the container runs with unbounded memory and can consume as much RAM as available on the host, preventing the OOM (Out of Memory) killer from terminating it even when it allocates excessive memory.

A Dockerfile at /home/interview/Dockerfile successfully builds a Go application, but the resulting Docker image is over 800MB in size due to including the full Go toolchain, build dependencies, and source code. Production images must be under 200MB.

You have a container image myapp:fs built from /home/interview/Dockerfile that performs intensive filesystem write operations. Docker is currently using the overlayfs storage driver and write performance is a bottleneck.

You have a Git repository at /home/interview/repo where you've just completed version 3.1.0 of your application. You need to create an annotated tag to mark this release with a proper message and metadata.

The repository at /home/interview/repo is in a broken state. Local and remote branches have diverged with no common ancestor. Consequently, git push origin main fails with a non-fast-forward error, and git pull origin main fails because the histories are unrelated.

You have two separate Git repositories at /home/interview/repo-a (5 commits) and /home/interview/repo-b (4 commits) developed independently. Create a new monorepo at /home/interview/monorepo that combines both repositories into separate subdirectories using subtree (project-a/ and project-b/) while preserving the full commit history from both repositories.

You have a Git repository at /home/interview/repo where you were in a detached HEAD state, made 3 commits, then switched back to the main branch. Those 3 commits are now unreachable and appear to be lost since no branch references them.

A file named secrets.env containing sensitive credentials exists in the repository's commit history. You need to purge this file from the entire history.

You have a Git repository at /home/interview/repo where the config.js file has been modified in the last two commits, but those changes introduced bugs. You need to restore only config.js to the version it had 2 commits ago without affecting any other files

Uncommitted changes on feature-ui prevent you from switching branches to fix a critical bug on main.

You have a Git repository at /home/interview/repo where you are working on a new feature on the feature-login branch with uncommitted changes in login.js. A fix is needed on the dev branch: app.js contains a JavaScript syntax error that can be identified by running node app.js.

You have a Git repository at /home/interview/repo that contains a submodule in the vendor/utils directory. The submodule is pointing to an old commit, but newer commits exist on the submodule's remote repository.

Your application needs to react to configuration changes without requiring a pod restart.

You have a Kubernetes cluster where the deployment webapp in namespace prod is stuck in CrashLoopBackOff.

The application exposes a health check endpoint at /healthz on port 8080. The container image already includes a config directory with the required files, but the deployment configuration has issues preventing the pod from starting correctly.

Because the config directory already exists and is used by the application, mounting a ConfigMap over the entire directory would overwrite it and cause the app to fail. The deployment needs to be fixed so configuration is injected without replacing the existing directory.

A CronJob named cleanup in the ops namespace is failing to trigger as expected. It has an incorrect schedule, relies on the default timezone (which may not match the server), and retains too many completed jobs, cluttering the history.

You need to extend the Kubernetes API to support a proprietary resource type called "Widget".

In namespace disco, the application discovery-app relies on DNS-based discovery of peer pods. The service discovery-svc is supposed to return all backend pod IPs, but currently:

kubectl exec deploy/testpod -n disco -- nslookup discovery-svc.disco.svc.cluster.local

returns only one IP instead of individual pod IPs. The application cannot discover its peers and fails to form a cluster.

You have a Kubernetes cluster where a job named data-loader in namespace ops fails with permission denied when calling the Kubernetes API. The job manifest file is located at /home/interview/data-loader-job.yaml.

A Deployment named backend in the dev namespace is failing to start. Pods are stuck in ImagePullBackOff due to configuration issues.

You are deploying a distributed database named dns-app in the dev namespace. This application requires that each Pod be addressable by a predictable, unchanging hostname (e.g., dns-app-0.dns-app) so the nodes can find each other for data replication.

Two teams share a cluster and require strict isolation with specific exceptions for inter-team communication.

A memory-intensive application named oom-demo is repeatedly crashing. The pod status shows CrashLoopBackOff, but you need to confirm the underlying cause is an "Out Of Memory" (OOM) error and fix it by increasing the memory limit.

An application requires TLS certificates for internal service communication.

The cluster has no StorageClass configured for dynamic volume expansion.

Your team has a stable deployment app-v1 running in namespace canary. They want to test a new version (v2) with approximately 1/3 of traffic going to v2 and 2/3 remaining on v1.

Log rotation has stopped working correctly, and you suspect that /var/log might be mounted on a different filesystem with limited space or incorrect mount options.

The developer account dev has been locked out of the server. Security logs indicate the SSH daemon's authentication failure limit was triggered.

One of your long-running Node.Js services (process name node) has been slowing down over several hours of uptime. CPU usage is normal, disk I/O is normal, but the server is gradually running out of memory.

You have noticed an unexpected spike in system load. You suspect a batch of recently spawned jobs is responsible and need to isolate processes that started within the last few minutes.

Your server cannot create new files. Commands like touch fail with "No space left on device" errors, but df -h shows plenty of free disk space. The filesystem has exhausted available inodes.

The server is consuming excessive resources. This server is used by multiple teams with their own credentials (e.g. each team has a username dev-team, qa-team, ops-team etc).

You're troubleshooting a service that produces untagged log output when run manually, making it difficult to analyze timing and sequence of events.

Each application environment (staging, dev, prod) has its own configuration file stored under /etc/app/envs/, and each file currently has multi_az = false and availability_zone = "us-east-1a". Manually editing each file is error-prone and inefficient, so the change must be automated.

Your application uploads files from /tmp/app/, but the maximum allowed file size is 1 MB, and some files exceed this limit.

The server has unmounted partitions that are not being used and could be utilized for additional storage.

A web-scraper systemd service is running on this system, making continuous HTTP requests. The service has started experiencing connection failures - logs show HTTP status 000 errors, indicating connections cannot be established even though the network is functional and remote servers are accessible.

A service on your server is running on port 8080, but you now need it to also be reachable on port 8081. The application cannot be restarted and its configuration cannot be changed.

You suspect the web service isn't receiving HTTP requests, and you need to confirm network traffic to port 80.

Your server uses multiple network interfaces and may have incorrect routing for a specific subnet. You need to verify and fix it to ensure proper traffic flow.

A minimal HTTPS webserver script (webserver.sh) listening on port 8443 fails to establish secure connections. The bundled SSL certificate (old_server.crt) lacks a Subject Alternative Name (SAN) for the local IP 127.0.0.1, causing hostname verification failures.