Why this matters
- This topic directly affects how reliably BookStore reaches production — environments are promotion stages with parity rules.
- Interviewers connect hands-on commands and manifests to real delivery stories, not buzzwords.
- Later modules assume you can explain both the why and the concrete file or command involved.
- Platform maturity shows up when teams automate this instead of relying on tribal knowledge.
Rolling
v1.0 pods3 running
v1.1 pod addedOne at a time
v1.0 pods removedZero downtime
Blue/Green
Blue (v1.0)Live traffic
Green (v1.1)Validated, idle
Switch trafficInstant cutover
Standard environments
Developers iterate locally with Compose. Staging runs the same Kubernetes chart as production with smaller replicas and anonymised data.
Key ideas
- Development — fast feedback, local dependencies
- Staging — production-like networking and config shape
- Production — real traffic and strict change control
Promotion flow
BookStore auto-deploys main to staging on green CI. Production promotion requires approval, synthetic checkout test, and comparison of error budget burn.
Key ideas
- Gates — automated tests plus human approval
- Parity — same Helm chart, different values
- Config drift — detect with GitOps diff
- Rollback — redeploy previous image digest
Configuration parity
Environment differences live in ConfigMaps, Secrets, and Helm values — never in divergent Dockerfiles.
Key ideas
- SPRING_PROFILES_ACTIVE — profile per env
- Secrets — injected at deploy, not baked in images
- Data — synthetic orders in staging
- Feature flags — decouple deploy from release
apiVersion: v1
kind: ConfigMap
metadata:
name: bookstore-config
namespace: staging
data:
SPRING_PROFILES_ACTIVE: staging
LOG_LEVEL: debug
Quick recall
Everything you need if you only revisit this box.
- BookStore uses environments and promotion as a standard delivery practice.
- Prefer automation and versioned config over manual server changes.
- Staging proves changes before customer-facing promotion.
- Observability confirms success — do not rely on silence alone.
- Rollback plans must be tested, not invented during an outage.
- Security and least privilege apply to every pipeline and cluster role.
Test yourself
Answer these before moving on — recall is what makes it stick.