Skip to content

K3s Fixture

Summary

The repository-owned fixture under examples/k3s-e2e/ is Meridian's real-cluster regression target for Meridian itself.

Why it exists

Meridian needs one authoritative cluster-level validation environment that the project owns end to end. Using a repo-owned fixture keeps the acceptance surface small, editable, and understandable.

What it includes

The fixture uses four application roles:

  • storefront
  • checkout
  • inventory
  • trafficgen

These services are intentionally limited so regressions can be diagnosed from artifacts instead of from a large demo stack.

Scenario model

Scenarios include:

  • happy
  • drop-traces
  • misroute-logs
  • auth-fail
  • backend-unreachable

Negative scenarios are expected to pass when the expected failure mode is detected correctly.

Acceptance signals

Primary signals on the project VM are:

  • pod logs
  • Prometheus evidence
  • gateway counters

Tempo and Loki remain supporting evidence, but are not the sole blocking signal in every scenario because they are not always fully run-scoped.