AWS SysOps and CloudOps Scenario Based Interview Questions: Troubleshooting, Reliability, Cost and Security Cases

AWS SysOps Associate 60 min min read Updated: Nov 17, 2026 Advanced
AWS SysOps and CloudOps Scenario Based Interview Questions: Troubleshooting, Reliability, Cost and Security Cases
Advanced Topic 2 of 3

How to Handle Scenarios

  1. Clarify: what changed, what is the impact, since when, how many users?
  2. Stabilize first (rollback, scale up, failover) then find the root cause
  3. Check in order: recent changes, health and metrics, logs, dependencies, network, permissions
  4. Say what you would check and why. Name AWS tools (CloudWatch, CloudTrail, Config, Reachability Analyzer, Performance Insights)
  5. Close the loop: fix, prevent, alert, document

Troubleshooting Cases

Q1. Users report the website is down right after a deployment. What do you do?

Check the ALB metrics (5xx, healthy hosts) and target health first. If targets are unhealthy after the release, roll back (previous AMI or blue/green swap back), then read application logs and the deployment record to find why health checks fail. Afterward add canary deployment, rollback alarms and a health endpoint that checks dependencies.

Q2. An EC2 instance is running but you cannot SSH to it.

Check instance status checks. Then security group (port 22 from your IP), NACL both ways, route to an Internet Gateway and public IP, key and user name. If network is fine, use Session Manager or the serial console. If the OS is broken, detach the root volume, attach it to a rescue instance, fix the config, and reattach.

Q3. RDS CPU is at 95 percent and the app is slow.

Open Performance Insights to see top SQL and wait events. Add or fix indexes, kill runaway queries, check connection counts and recent deployments. As quick relief scale up the instance or add a read replica for reporting. Add alarms and review the slow query log.

Q4. S3 returns Access Denied to an application running on EC2.

Confirm the role the instance really uses (sts get-caller-identity), then check identity policy, bucket policy, explicit denies, SCPs, Block Public Access, KMS permissions for SSE-KMS objects and VPC endpoint policy. Use CloudTrail errorMessage and the policy simulator.

Q5. Auto Scaling did not add instances during a traffic spike.

Check max size, quotas (vCPU limit), scaling policy and alarm state, cooldown or warm-up, suspended processes, failed launches (describe-scaling-activities), AZ capacity or subnet IP exhaustion, and launch template errors such as a KMS key blocking an encrypted AMI.

Q6. A disk filled up on a production server at night.

Alarm should have fired from the CloudWatch agent metric. Free space safely (logs, temp files) using an SSM runbook, then grow the volume and file system online. Prevent with log rotation, alarms at 80 percent, EBS autoscaling automation, and retention on logs.

Q7. A CloudFormation stack update failed and rolled back. How do you investigate?

Read describe-stack-events for the first FAILED resource and reason. Common causes are permissions, quotas, name conflicts or a bad property. Fix and re-run through a change set. If it is UPDATE_ROLLBACK_FAILED, fix the manual change and continue-update-rollback.

Q8. An application cannot reach an RDS database after a change.

Run Reachability Analyzer from the app ENI to the database ENI, then check database security group rules (prefer referencing the app SG), subnet routes, NACLs, DNS endpoint, and whether the instance is available. Check Flow Logs for REJECT.

Q9. Lambda is throttled and messages are failing.

Check Throttles, concurrency (account and reserved), and event source settings. Raise limits or add reserved concurrency, put SQS in front with a DLQ and backoff, and make downstream systems handle the load.

Q10. Website is slow only for users far from the Region.

Use CloudFront for caching, Route 53 latency routing or Global Accelerator, and check TLS and origin latency. Validate with CloudWatch RUM or synthetics.

Q11. Logs are missing in CloudWatch.

Check agent status and config, IAM permissions (logs:PutLogEvents), network (NAT or logs endpoint), file path and rotation, KMS key policy, correct Region and time range.

Reliability and Design Cases

Q1. Design a web app that survives an AZ outage.

ALB across 3 AZs, ASG across 3 AZs with spare capacity, one NAT gateway per AZ, Aurora or RDS Multi-AZ, ElastiCache with replicas, S3 for assets. Test by stopping instances in one AZ with Fault Injection Service.

Q2. RTO is 1 hour and RPO is 5 minutes across Regions. What do you design?

Pilot light or warm standby: Aurora Global Database or cross-Region replica, AMIs and templates in the DR Region, replicated secrets and keys, Route 53 failover, and a rehearsed runbook plus quota increases in the DR Region.

Q3. You must prove backups are restorable for an auditor.

Use AWS Backup restore testing on a schedule, keep reports, store evidence in Audit Manager, and run DR drills documenting the measured RTO.

Q4. How do you patch 300 servers with minimal risk?

Tag servers, use patch baselines with a delay, scan first, install in maintenance windows with rate control (dev first), snapshot before prod, verify health after, and report compliance. For auto scaled fleets bake patched AMIs and do an instance refresh.

Q5. A single NAT gateway failed and several AZs lost internet access.

The design had a single point of failure. Create a NAT gateway per AZ and route each private subnet to its local NAT. Use endpoints to reduce dependency.

Q6. How do you move from manual fixes to automation safely?

Write runbooks in SSM Automation, test in dev, start in notify-only mode, add approvals, scope by tags, rate limit, log everything, then enable auto-remediation for low-risk fixes.

Cost Cases

Q1. The AWS bill jumped 40 percent this month. What do you do?

Use Cost Explorer grouped by service, account, Region and tag, and check Cost Anomaly Detection root causes. Typical causes: NAT data processing, cross-AZ or internet data transfer, a forgotten large instance, a log flood, S3 requests or replication. Fix the cause and add budgets and anomaly alerts.

Q2. How do you reduce cost without hurting performance?

Remove idle resources, right-size with Compute Optimizer using p95 metrics, schedule non-production, move to gp3 and Graviton, use S3 lifecycle and Intelligent-Tiering, then buy Savings Plans for the steady baseline and use Spot for tolerant work.

Q3. Teams do not tag resources. What do you do?

Define a tag standard, enforce with tag policies, SCPs and Config required-tags, add tags in IaC modules, activate cost allocation tags and report untagged spend weekly.

Q4. Data transfer costs are high between services. Ideas?

Keep chatty services in the same AZ when possible, use VPC endpoints (S3 gateway endpoint is free), CloudFront for egress, compress data, and review cross-Region replication.

Security Cases

Q1. GuardDuty reports a high severity finding on an EC2 instance. What do you do?

Triage (finding type, instance, time). Contain by moving it to a quarantine security group and removing or revoking its role sessions. Preserve evidence with EBS snapshots and logs. Investigate with CloudTrail and Detective, eradicate (rebuild from a clean AMI, rotate secrets), recover and write a post-incident review.

Q2. A public S3 bucket was found. What now?

Turn on Block Public Access (account level too), check what data was exposed (Macie, server access logs, CloudTrail data events), fix the cause with Config rule plus auto-remediation, and add an SCP and alerting.

Q3. How do you give a contractor temporary access?

Use IAM Identity Center or a role with limited permissions, a time-limited session, MFA, and ExternalId for third-party accounts. Remove access when done.

Q4. Prevent anyone from disabling CloudTrail across all accounts.

Use an organization trail in a log archive account and an SCP denying cloudtrail:StopLogging and DeleteTrail. Add Config rules and alerts on changes.

Q5. The application stores a database password in code. How do you fix it?

Move it to Secrets Manager with rotation, grant the role access, remove it from code and history, rotate the leaked password, and add secret scanning to the pipeline.

Conclusion

In scenario questions, stabilize first, then investigate in a clear order using the right AWS tools, explain trade-offs, and finish with prevention: alarms, automation, runbooks, IaC and tests.

Quick Summary: troubleshooting; reliability; cost; security.

Get Newsletter

Subscibe to our newsletter and we will notify you about the newest updates on Edugators