serve: autoscaler `latest_version` resets on controller restart, causing scale churn · skypilot-org/skypilot#8562

2026-01-13T19:00:43.000Z

### Summary - After Serve controller restarts, autoscaler latest_version stays at INITIAL_VERSION (1) while replicas are launched at the latest service version (e.g., 3). - This mismatch causes the autoscaler to repeatedly scale up and then immediately scale down the newly launched replicas, even at 0 RPS. ### What we saw - One of our service deployments had a constant loop: autoscaler requests 3 scale‑ups (min=2 + overprovision=1), then scales down the same new replicas. - The loop starts right after the controller restarts. - DB state (example): - services shows current_version=1 but active_versions=[3]. - version_specs only contains version=3. - Logs show: - Requests per second 0.0, target replicas computed as 0 or 2, but scale‑ups still requested. - Immediate scale‑down of the replicas that were just launched. ### Repro 1. Launch a service and update it to version > 1. 2. Restart the controller (or let K8s reschedule it). 3. Observe autoscaler logs: continuous scale up/down churn, even at 0 RPS. I can provide a suggestion/fix in a bit.

(3 comments) (0 reactions) (0 assignees)Python (311 forks)batch import

good first issuegood starter issues

Métricas do repositório

Stars: (4.859 stars)
Métricas de merge de PR: (Mesclagem média 2d 23h) (146 fundiu PRs em 30d)

Description

Summary

After Serve controller restarts, autoscaler latest_version stays at INITIAL_VERSION (1) while replicas are launched at the latest service version (e.g., 3).
This mismatch causes the autoscaler to repeatedly scale up and then immediately scale down the newly launched replicas, even at 0 RPS.

What we saw

One of our service deployments had a constant loop: autoscaler requests 3 scale‑ups (min=2 + overprovision=1), then scales down the same new replicas.
The loop starts right after the controller restarts.
DB state (example): - services shows current_version=1 but active_versions=[3]. - version_specs only contains version=3.
Logs show: - Requests per second 0.0, target replicas computed as 0 or 2, but scale‑ups still requested. - Immediate scale‑down of the replicas that were just launched.

Repro

Launch a service and update it to version > 1.
Restart the controller (or let K8s reschedule it).
Observe autoscaler logs: continuous scale up/down churn, even at 0 RPS.

I can provide a suggestion/fix in a bit.

Guia do colaborador

Direção de pesquisa: Investigue a inicialização do latest version do autoscaler e garanta que ele persista após a reinicialização do controlador.
Pilha de tecnologia: python
Domain: backenddevops
Tipo Issue: Bug
Difficulty: 3
Tempo estimado: Meio dia
Status da atividade: Ativo
Clarity: Claro
Prerequisites: PythonKubernetes
Simpatia para novatos: 60