MIR-993: Configurable port-wait timeout for slow-cold-init addons

[evanphx]

Summary

• Fixes MIR-993: MySQL addon sandboxes consistently fail their first-boot health check on loaded dev hardware because MySQL 8's cold init (--initialize → temporary setup server → real server) takes ~20s, exceeding the hardcoded 15s port-bind budget. The pool then marks the sandbox DEAD and crash-backs-off, so the fast (~1s) second boot is delayed too.
• Adds an optional port_wait_timeout to SandboxSpec, threaded through both the legacy controller path (controllers/sandbox/sandbox.go) and the saga create path (controllers/sandbox/create_saga.go — via a new PortWaitTimeout field on bootContainersOut → waitPortsIn, wired by matching saga:"port_wait_timeout" tags). Empty/invalid values fall back to the existing 15s default so normal apps and typos are unaffected.
• Sets budgets on the affected addons via a new PortWaitTimeout field on addon.CreateSandboxPoolSpec:
  • MySQL: 60s (shared + dedicated)
  • Postgres: 60s (shared + dedicated)
  • RabbitMQ: 90s (dedicated) — Erlang VM + mnesia boot is slower than SQL engines
• specsMatch now compares PortWaitTimeout so changing the budget on an existing pool is detected as a spec drift.

Test plan

• make lint — 0 issues
• Unit test TestResolvePortWaitTimeout covers empty / valid / zero / negative / garbage / bare-number fallback behavior
• Saga tests TestCreateSandboxSaga_PortWaitTimeoutDefault / _Override assert the spec value flows through to WaitForPort
• TestSpecsMatchCoversAllFields and TestSandboxControllerFrozen hashes updated; saga path carries the equivalent change
• Manual repro in local dev cluster: server log on the MySQL shared pool sandbox shows waiting for ports to be bound ... timeout: 1m0s, first boot comes up cleanly without a DEAD / crash-backoff cycle
• make test-blackbox — existing MySQL addon blackbox still green

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: 46064378-17a4-40b0-93eb-1082487878f1

📥 Commits

Reviewing files that changed from the base of the PR and between c79f8d2 and 310987c.

📒 Files selected for processing (15)

• api/compute/compute_v1alpha/schema.gen.go
• api/compute/schema.yml
• controllers/deployment/launcher.go
• controllers/deployment/specs_match_test.go
• controllers/sandbox/create_saga.go
• controllers/sandbox/create_saga_test.go
• controllers/sandbox/port_wait_timeout_test.go
• controllers/sandbox/sandbox.go
• controllers/sandbox/sandbox_frozen_test.go
• pkg/addon/framework.go
• pkg/addon/mysql/dedicated.go
• pkg/addon/mysql/shared.go
• pkg/addon/postgresql/dedicated.go
• pkg/addon/postgresql/shared.go
• pkg/addon/rabbitmq/dedicated.go

✅ Files skipped from review due to trivial changes (5)

• pkg/addon/postgresql/shared.go
• controllers/deployment/specs_match_test.go
• pkg/addon/mysql/shared.go
• api/compute/schema.yml
• pkg/addon/postgresql/dedicated.go

🚧 Files skipped from review as they are similar to previous changes (6)

• pkg/addon/mysql/dedicated.go
• pkg/addon/rabbitmq/dedicated.go
• controllers/sandbox/sandbox_frozen_test.go
• controllers/deployment/launcher.go
• controllers/sandbox/port_wait_timeout_test.go
• controllers/sandbox/sandbox.go

---

📝 Walkthrough

Walkthrough

Added a new optional PortWaitTimeout string field to SandboxSpec and schema (duration parsed via time.ParseDuration). Controller logic was updated to use resolvePortWaitTimeout with a 15s default instead of a hardcoded timeout; specsMatch compares the new field. Saga inputs/outputs and wait logic were adjusted, tests added/updated (including port-wait unit tests, mock runtime, fingerprint and frozen-file hash updates). The addon framework exposes PortWaitTimeout, and MySQL/PostgreSQL pools set 60s while RabbitMQ sets 90s.

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Actionable comments posted: 1

🧹 Nitpick comments (1)

controllers/deployment/launcher.go (1)

846-848: Prefer semantic timeout comparison over raw string equality.

Line 846 currently mismatches equivalent duration strings (for example, 60s vs 1m), which can trigger unnecessary pool replacement.

♻️ Suggested refactor

-	if spec1.PortWaitTimeout != spec2.PortWaitTimeout {
-		return fmt.Sprintf("port wait timeout mismatch: %s vs %s", spec1.PortWaitTimeout, spec2.PortWaitTimeout), false
-	}
+	resolve := func(raw string) time.Duration {
+		if strings.TrimSpace(raw) == "" {
+			return 15 * time.Second
+		}
+		d, err := time.ParseDuration(raw)
+		if err != nil || d <= 0 {
+			return 15 * time.Second
+		}
+		return d
+	}
+	t1 := resolve(spec1.PortWaitTimeout)
+	t2 := resolve(spec2.PortWaitTimeout)
+	if t1 != t2 {
+		return fmt.Sprintf("port wait timeout mismatch: %s (%s) vs %s (%s)",
+			spec1.PortWaitTimeout, t1, spec2.PortWaitTimeout, t2), false
+	}

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@controllers/deployment/launcher.go` around lines 846 - 848, The current
string equality check for PortWaitTimeout on spec1 and spec2 can misreport
equivalent durations (e.g. "60s" vs "1m"); change the comparison to parse both
spec1.PortWaitTimeout and spec2.PortWaitTimeout with time.ParseDuration, handle
parse errors (log/return a clear mismatch/error if parsing fails), and compare
the resulting time.Duration values numerically so equivalent durations are
treated as equal; reference the symbols spec1.PortWaitTimeout and
spec2.PortWaitTimeout and ensure the mismatch message reflects the original
strings or the parsed durations for clarity.

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Inline comments:
In `@api/compute/schema.yml`:
- Around line 26-32: Update the port_wait_timeout documentation to state that
empty, unparsable, or non-positive durations (<=0, e.g. "0s") are treated the
same by the controller and will fall back to the default of 15s; reference that
values are parsed via time.ParseDuration and give an example of a non-positive
value being ignored so users don't assume "0s" disables the wait. Keep the
existing examples (e.g. "60s") and explicitly note addons needing longer waits
should set a larger positive duration.

---

Nitpick comments:
In `@controllers/deployment/launcher.go`:
- Around line 846-848: The current string equality check for PortWaitTimeout on
spec1 and spec2 can misreport equivalent durations (e.g. "60s" vs "1m"); change
the comparison to parse both spec1.PortWaitTimeout and spec2.PortWaitTimeout
with time.ParseDuration, handle parse errors (log/return a clear mismatch/error
if parsing fails), and compare the resulting time.Duration values numerically so
equivalent durations are treated as equal; reference the symbols
spec1.PortWaitTimeout and spec2.PortWaitTimeout and ensure the mismatch message
reflects the original strings or the parsed durations for clarity.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: 056a84f3-60d3-4147-9e40-1c09ea2ed031

📥 Commits

Reviewing files that changed from the base of the PR and between a29b71d and c79f8d2.

📒 Files selected for processing (15)

• api/compute/compute_v1alpha/schema.gen.go
• api/compute/schema.yml
• controllers/deployment/launcher.go
• controllers/deployment/specs_match_test.go
• controllers/sandbox/create_saga.go
• controllers/sandbox/create_saga_test.go
• controllers/sandbox/port_wait_timeout_test.go
• controllers/sandbox/sandbox.go
• controllers/sandbox/sandbox_frozen_test.go
• pkg/addon/framework.go
• pkg/addon/mysql/dedicated.go
• pkg/addon/mysql/shared.go
• pkg/addon/postgresql/dedicated.go
• pkg/addon/postgresql/shared.go
• pkg/addon/rabbitmq/dedicated.go