Spread replicas across nodes #393

Parent

Rewrite the spec and the compiler in place, release 1.0.0, and hand over every Project

What to build

An Application author writes across: node on replicas, and no two replicas share a node. A rollout still completes when replicas equal eligible nodes. Settled in Decide how the model meets soft scheduling, spread, anti-affinity and parked replicas (items 4 and 5) and Decide which conditions on the settled decisions change the model (decision 1), where a drill confirmed both the stall and the fix on a plain Deployment.

Acceptance criteria

• replicas: {count, reason, across: node} renders required pod anti-affinity on the hostname, scoped to one revision. across admits node only.
• Fewer eligible nodes than count is E_PLACEMENT_UNSATISFIABLE.
• Without across, two replicas on one node stay legal.
• Chapter 18 states the cost: while a rollout runs, an old and a new pod may share a node.
• The result of Drill a Flagger promotion of a Process spread across nodes is read first. If the promotion fails, the choice between the revision key and a rollout without surge reopens with the maintainer.
• The chapter states the rule and its refusals, with short rationale only, and CONTEXT.md carries each new term in the same pull request.
• A worked example shows it; its resolved artifact and its rendered tree are updated and reviewed by hand, never regenerated in CI.
• Each refusal has a refusal example with its diagnostics file.
• The JSON Schemas and their corpus admit the new shape and refuse a malformed one.
• The behaviour-ledger row lands with its test, and npm run verify passes.

Blocked by

• Cut what the map cut and release package 1.0.0
• Drill a Flagger promotion of a Process spread across nodes