josh/electric-cherry
1
0
Fork 0
Code Issues 18 Pull Requests Actions 1 Packages Projects Releases Wiki Activity

Finish autoscaling diagnostics and document pilot rollout

Browse Source
This commit is contained in:
josh
2026-10-06 23:18:39 -04:00
parent e271f31325
commit a61eccaed6
10 changed files with 135 additions and 26 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/lobby-autoscaling-plan.md
+11 -1
View File
@@ -1,6 +1,6 @@
# Lobby autoscaling implementation plan
Status: Approved October 6, 2026. Implementation is underway; live rollout remains gated on the dry run and pilot below.
Status: Approved and implemented October 6, 2026. Controller support and agent 1.7.0 are complete; production dry run, pilot measurements, and live rollout remain pending. See [usage and the pilot procedure](autoscaling.md).
Introduce an opt-in policy for each game and map variant that keeps a small number of empty, ready lobbies available. When the last spare fills, request another lobby through Provisioning. Remove surplus lobbies only after sustained, confirmed inactivity. This reduces idle hosting while preserving room for new groups, including when existing Zombies sessions have free player slots but are already well into a game.
@@ -136,3 +136,13 @@ Run targeted Vitest coverage for planner, collections, provisioning, cancellatio
Provide a global pause as well as per-policy pause. It stops new decisions and withdraws unstarted work using the same cancellation contract. Keep the controller and agent running to settle delivered operations and cleanup; do not delete policies or reservations to bypass recovery.
After outstanding operations settle, release map ownership. Operators can restore a fixed lineup through Collections using normal capacity and empty-server gates. Turning autoscaling off does not recreate removed lobbies or remove lobbies it created. Retain decision and Provisioning history for review, and keep the additive schema compatible with all policies Off.
## Implementation and validation record
The implementation adds the pure evaluator and persistent policy model, dry-run requester and UI, exclusive map ownership with capacity-aware creation, and revocable removal authorization. It creates no production policies. The authorization claim is the final cancellation boundary; agents invalidate an occupied or expired attempt and require a new decision rather than retrying it later. The local claim window is five seconds, measured from the start of the authorization request, with another occupancy check before mutation.
Validation includes planner and snapshot tests, duplicate-admission and stale-preview tests, collection exact-match conflicts, operator cancellation, authorization expiry and host identity, controller/agent occupancy races, and agent restart recovery. Agent packaging publishes version 1.7.0 with the new capability.
The full controller suite reported 1,934 passing tests and two existing failures in `server-detail-api.test.ts` (plugin classification) and `watchdog.test.ts` (an occupied-telemetry RCON expectation). Both failures reproduce independently at the pre-implementation commit `a22461a`. They are not autoscaling regressions. Targeted controller tests, all 228 Python agent tests, type checking, and the production build passed.
An isolated Firefox review exercised policy creation, preview, save, keyboard focus, and 1440, 390, and 320 pixel layouts. A narrow-layout heading issue found during review was corrected. The preview used a temporary database and no production host credentials. Production CPU savings, external joinability, spare targets, and acceptable deficit durations still require the two-week observation period and operator-run pilot.