Cheat sheet + 20 flashcards covering CI/CD, Docker, Kubernetes, IaC, observability, deployment strategies, and incident management.
CI: merge frequently, test every commit. CD: code that passes CI goes automatically to staging (Delivery) or production (Deployment).
Packages app + dependencies into a portable, isolated container. Solves "works on my machine." Lighter than VMs — shares the host OS kernel.
Image: read-only layered snapshot (built from Dockerfile)
Container: running instance of an image
Automates deployment, scaling, self-healing, and load balancing of containers across a cluster.
Pod: smallest unit — one or more containers, shared network
Deployment: manages replicas, rolling updates
Service: stable network endpoint for a set of pods
Define infrastructure in code files — version controlled, repeatable, reviewable.
Terraform: declarative — provisions cloud resources (VMs, DBs, networks). Defines desired end state.
Ansible: imperative — configures existing machines via playbooks.
Metrics: time-series numbers (CPU %, request rate, error count). Prometheus + Grafana.
Logs: timestamped event records. Structured JSON logs → ELK / Loki.
Traces: end-to-end request journeys across services. Jaeger / Zipkin.
RED: Rate (req/s), Errors (error rate), Duration (latency p50/p95/p99)
SLI: metric measuring health (e.g. p99 latency)
SLO: target value (e.g. p99 < 200ms)
SLA: contractual commitment — violating it costs money
GitFlow: long-lived feature/develop/release branches — structured but slow, merge conflicts accumulate.
Trunk-based: everyone commits to main frequently using feature flags to hide incomplete work — faster, simpler, CI/CD-friendly.
Blue/Green: two identical envs — deploy to Green, test, flip traffic instantly. Rollback = flip back. Zero downtime.
Canary: roll out to 5% of users first, monitor errors, increase gradually. Issues affect only a fraction before full release.
1. Flow: fast, smooth delivery from dev to production — eliminate handoffs and bottlenecks.
2. Feedback: fast feedback loops at every stage — monitoring, testing, alerts.
3. Continual Learning: experiment, learn from failures, share knowledge.
Move testing, security, and quality checks earlier in the development lifecycle — closer to where code is written.
Defects found early are 10–100× cheaper to fix than in production.
Practices: unit tests in CI, SAST scanning, code review, linting on commit.
Liveness probe: is the container alive? Fail → K8s restarts it.
Readiness probe: is the container ready for traffic? Fail → K8s removes it from the load balancer without restarting.
Both prevent traffic going to broken or starting-up pods.
A blameless analysis after an incident: what happened, why, how to prevent recurrence.
Sections: timeline, root cause, contributing factors, action items with owners and due dates.
"Blameless" = focus on system/process failures, not individuals.
Space flip · → next · ← prev · G got it · R review