Phase 3 - SRE And Operations

This phase shifts from "how to deploy" to "how to run production systems safely."

Goal

By the end of this phase, you should be able to detect issues early, troubleshoot methodically, stabilize production systems, and explain the reasoning behind your actions.

Study Order

1. ../../05_Observability_and_Troubleshooting/Monitoring/README.md

2. ../../07_Interview_Preparation/devops-interview-playbook.md

3. ../../07_Interview_Preparation/general-interview-questions.md

4. ../../07_Interview_Preparation/interview-questions-medium.md

5. ../../07_Interview_Preparation/interview-questions-hard.md

6. ../../04_Infrastructure_as_Code_and_Cloud/Cloud_Services/azure-medium-questions.md

7. ../../04_Infrastructure_as_Code_and_Cloud/Cloud_Services/azure-hard-questions.md

What To Master

Observability

• metrics, logs, and traces

• Prometheus, Grafana, and alerting

• SLI, SLO, SLA, and error budgets

• high-cardinality risks and noisy alerts

Troubleshooting

• using metrics before guessing

• events, logs, rollout history, and node inspection

• common Kubernetes and Terraform failure states

• how to distinguish app issues from infrastructure issues

Reliability

• rollback strategy

• canary, blue-green, and rolling trade-offs

• graceful degradation and load shedding

• incident mitigation versus long-term prevention

Security And Operational Hygiene

• secrets handling

• least privilege

• image and dependency scanning

• policy as code and runtime controls

Cost And Capacity

• overprovisioning versus reliability margin

• autoscaling logic

• rightsizing

• cloud waste and idle resources

Hands-On Tasks

1. Build a Grafana dashboard and define one actionable alert.

2. Simulate a failing deployment and practice rollback.

3. Debug a CrashLoopBackOff or Pending pod in a lab.

4. Write a short runbook for a high-latency service.

5. Write a short RCA for a deployment or config incident.

Checkpoint Questions

• How do logs, metrics, and traces work together during an outage?

• Why is symptom-based alerting better than paging on every CPU spike?

• What do you check first in CrashLoopBackOff?

• If HPA scales pods but latency stays high, what does that imply?

• What is the difference between immediate mitigation and permanent fix?

Exit Criteria

Move to Phase 4 only when you can:

• explain SLO-based thinking clearly

• investigate a production-style issue in a structured way

• name the commands you would use for Kubernetes and host troubleshooting

• describe a safe rollback strategy

• talk about monitoring, security, and cost as one operating model rather than separate topics