Operate the platform itself as a reliable product: define SLOs, on-call and support ownership, manage intake/roadmaps/deprecations, measure adoption and task success, align platform/enabling/application teams, and evolve the platform from evidence.
If repository creation, environment provisioning or deployment orchestration is unreliable, the platform becomes a delivery dependency. Measure availability, latency and failure modes for the capabilities developers rely on.
High portal traffic may coexist with ticket-heavy work. Combine adoption with completion rate, lead time, retry rate, support escalation and developer feedback.
A template/API version cannot be changed as casually after many teams depend on it. Track consumers, announce deprecations, provide migrations and observe adoption before removing old paths.
Golden-path services create shared dependencies. Define who is paged when repository creation fails, when the GitOps control plane is degraded, when a shared cluster is unhealthy, and when only one application is broken. Platform teams should own the capabilities they provide, with documented escalation into infrastructure or application owners when the fault domain changes.
A platform team needs a visible intake mechanism, prioritization criteria, capacity for reliability work, and a roadmap tied to developer outcomes. Stakeholder demand, support volume, platform SLOs and adoption evidence should influence funding and sequencing; the loudest request should not automatically win.
Stream-aligned application teams should own their services. The platform team provides reusable paved-road capabilities. Enabling or specialist teams can help teams adopt new practices or supply deep expertise. The names are less important than clear ownership, collaboration modes and avoiding a permanent ticket-based dependency on the platform team.
AI workloads can add accelerator scheduling, model/artifact registries, inference endpoints, quota and cost controls, data/model provenance, and specialized observability. These are emerging platform capabilities, not a replacement for the core product principles in this course. Treat them as an extension after the general workload contract is sound.
request success rate · provisioning p95 · self-service completion · support escalation · golden-path adoption · developer feedbackGoal: Calculate success rate and identify whether a product target is met.
cat > /tmp/platform-slo.json <<'EOF'
{"requests":1000,"successes":985,"target":0.99,"median_minutes":3,"p95_minutes":11,"support_escalations":18}
EOFpython3 - <<'PY2'
import json
x=json.load(open('/tmp/platform-slo.json'))
rate=x['successes']/x['requests']
print(f'success={rate:.2%} target={x["target"]:.2%}')
assert rate < x['target']
assert x['p95_minutes'] >= x['median_minutes']
print('SLO miss detected')
PY2Use failure/escalation evidence to choose one platform improvement rather than adding an unrelated feature.
For three failures—one application error, one GitOps control-plane outage and one shared-cluster outage—write the primary pager, escalation target and evidence required at handoff.
Platform teams need the same reliability and product feedback discipline they ask application teams to adopt.
Review platform capability telemetry and turn it into an on-call/roadmap decision rather than a dashboard-only exercise.
printf '%s
' 'platform_request_success_rate' 'platform_request_duration_seconds' 'platform_request_failures_total{reason=...}' 'platform_capability_users'printf '%s
' 'application incident -> app team' 'platform capability incident -> platform on-call' 'shared infrastructure incident -> platform/infra escalation'printf '%s
' 'intake -> prioritize -> roadmap -> deliver -> observe -> feedback -> deprecate'Three concept checks plus one scenario check.
It is a dependency in delivery workflows, so its reliability and latency affect developer outcomes.
Usage alone does not prove that developers complete work efficiently or without support.
Known consumers, deprecation communication, migration path and evidence that migration is complete or acceptable.
No. Investigate self-service completion, failure reasons and why users still require manual support.
Progress is saved in this browser immediately. When you sign in to My Learning, it can also sync across devices.
Attempt the core lab and knowledge checks before marking the module complete.