feat(ecs): Batch 3 — services with rolling deployments

[vieiralucas]

Summary

Batch 3 of the ECS rollout. Adds the full services surface on top of the Batch 2 task execution: `CreateService`, `UpdateService`, `DeleteService`, `DescribeServices`, `ListServices`, `ListServicesByNamespace` with desired-count enforcement and rolling deployments.

Highlights:

• `CreateService` spawns tasks to match `desiredCount` and tags each with `startedBy=ecs-svc/` so the tasks reconcile back to the service.
• `UpdateService` scaling (up: spawn new tasks; down: SIGTERM excess via runtime + flip `desiredStatus=STOPPED` synchronously).
• `UpdateService` rolling deployment: new `taskDefinition` marks the previous PRIMARY deployment ACTIVE, creates a new PRIMARY, and drains tasks on the old revision while new ones spin up. `minimumHealthyPercent` / `maximumPercent` + deployment circuit breaker are honoured in `deploymentConfiguration`.
• `DeleteService` refuses while `desiredCount > 0` unless `force=true`; forced path scales to 0 and stops every task under the service.
• `DescribeServices` derives `runningCount` / `pendingCount` from live task state on each read.
• Snapshot schema bumped to v3 to persist services.

Deferred to Batch 4: task sets (EXTERNAL deployment controller), container instances, capacity providers, task protection, `ExecuteCommand`, EventBridge state-change events.

Test plan

• `cargo test -p fakecloud-e2e --test ecs` — 18 scenarios pass (4 new for services: create/describe/list round-trip, scale up+down, rolling deployment, force-delete)
• `cargo test -p fakecloud-conformance --test ecs` — 31 conformance tests pass (all 6 new actions have live Smithy-shape checksums)
• `cargo clippy --workspace --all-targets -- -D warnings` — clean
• `cargo fmt --all --check` — clean
• CI green on all checks
• Cubic review clean

---

Summary by cubic

Adds ECS services with rolling deployments on top of task execution, with desired-count enforcement and live running/pending counts. Also fixes an over-spawn issue when scaling down during a rolling deployment. Snapshot schema bumped to v3.

• New Features

  • Added CreateService, UpdateService, DeleteService, DescribeServices, ListServices, ListServicesByNamespace.
  • Scaling: spawns tasks on scale-up; on scale-down flips excess to desiredStatus=STOPPED and SIGTERMs via the runtime.
  • Rolling deployments: new taskDefinition creates a new PRIMARY, marks the previous PRIMARY ACTIVE, and drains old tasks; honors minimumHealthyPercent, maximumPercent, and the deployment circuit breaker.
  • Service linkage and reads: tasks tagged startedBy=ecs-svc/<name>; DescribeServices derives runningCount/pendingCount from live tasks; ListServices supports cluster/launchType/schedulingStrategy filters.

• Bug Fixes

  • Prevented over-spawning in UpdateService when scaling down and changing taskDefinition in the same call by computing replacements from the effective desired count (not from the total stop set).

^{Written for commit f7995e3. Summary will update on new commits.}

[codecov]

Codecov Report

❌ Patch coverage is 0.38168% with 783 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
crates/fakecloud-ecs/src/service.rs	0.00%	778 Missing ⚠️
crates/fakecloud-ecs/src/state.rs	37.50%	5 Missing ⚠️

📢 Thoughts on this report? Let us know!

[cubic-dev-ai]

1 issue found across 6 files

Prompt for AI agents (unresolved issues)

Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name="crates/fakecloud-ecs/src/service.rs">

<violation number="1" location="crates/fakecloud-ecs/src/service.rs:2195">
P1: Rolling deployment over-spawns tasks when combined with scale-down. `stop.len()` counts tasks stopped for both scaling *and* TD-drain, so new-task spawns exceed `effective_desired`. Use `effective_desired` as the target instead of `stop.len()`.</violation>
</file>

_{Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.}