feat(config): gate MySQL-only features for postgres targets

[Kiran01bm]

Summary

Gates deferred cutover — a MySQL-dialect feature — at apply request time with a typed, named-feature rejection instead of the generic no-engine failure that postgres targets previously fell through to. Also corrects Strata's advertised capability: it never executed deferred cutover, so it no longer claims it. MySQL and Vitess behavior is unchanged.

What

• pkg/schema/dialect.go: SupportsFeature(databaseType, feature) — deferred cutover is supported by MySQL and Vitess only; unknown database types and unknown features fail closed.
• pkg/api/plan_handlers.go: apply queueing rejects a deferred-cutover request against an unsupported target with a typed UnsupportedFeatureError, translated to HTTP 400 invalid_request naming the feature and database (previously a 500). Span status, apply metric, and triage log context are recorded on rejection.
• pkg/webhook: the apply PR comment renders the actionable named-feature message instead of a generic failure.
• Deliberately no config-time gate: pending drops is global server behavior, not a per-apply option — gating it at config validation would fail startup for existing mixed-type configs that never exercise it against postgres. Its gate belongs with the postgres engine work, where the feature is actually executed.

Why

Deferred cutover is a semantic of the Spirit/MySQL engine (gh-ost-style cutover deferral); it has no meaning for the planned Postgres engine's native path. Until the feature has a Postgres design of its own, the honest behavior is an explicit "not supported for this database type" at the earliest point the intent is visible — the apply request — not a generic routing error at execution time. The gate is additive: no engine exists for postgres yet, so nothing that works today changes.

Before                                  After

apply --defer-cutover                   apply --defer-cutover
  on postgres target                      on postgres/strata target
        │                                       │
        ▼                                       ▼
queue → execute → "no engine"           queue-time capability check
        → HTTP 500, generic error               │
        → PR comment: opaque failure            ▼
                                        UnsupportedFeatureError
                                        → HTTP 400 invalid_request,
                                          names feature + database
                                        → PR comment: actionable message

[Copilot]

Pull request overview

Adds explicit gating for MySQL-dialect-only features (pending drops, deferred cutover) so Postgres targets fail early with a named-feature error rather than falling through to a generic “no engine” style failure.

Changes:

• Introduces schema.SupportsFeature(databaseType, feature) and Feature identifiers to centralize dialect feature support checks.
• Adds validateDatabaseFeatures(...) and hooks it into server config validation and apply queueing.
• Expands tests to cover the new feature predicate and validation behavior.

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
pkg/schema/dialect.go	Adds feature identifiers and a predicate to check feature support per database type/dialect.
pkg/schema/dialect_test.go	Adds unit coverage for SupportsFeature across supported/unsupported database types/features.
pkg/api/plan_handlers.go	Adds apply-time gating for unsupported feature usage based on plan/request options.
pkg/api/config.go	Adds config-time validation helper to reject unsupported feature + database type combinations.
pkg/api/config_test.go	Updates config validation coverage and adds tests for validateDatabaseFeatures.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.