Files
yingqing/k8s/core/worker-deployment.yaml
T
seaislee1209andClaude Opus 4.8 03e7949307 feat(core): 模型选择框换 restraint chip 下拉 + worker 并发出图(修 -P solo 串行)
UI:脚本助手模型框从『原生 select + inline style』换成复用设计系统的 chip 下拉(向上展开,新增共享 .align-up 变体),去 inline、合规范。并发:本地 worker 从 -P solo(单任务串行=一次只生成一张)改 -P threads -c 4;生产 k8s worker --concurrency 2→4。基础资产/故事板/视频出图可并行。loading 动画基础设施(卡内+全局 spinner)已验证存在。

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-17 22:17:47 +08:00

44 lines
1.3 KiB
YAML

apiVersion: apps/v1
kind: Deployment
metadata:
name: airshelf-core-worker
labels:
app: airshelf-core-worker
spec:
replicas: 1
selector:
matchLabels:
app: airshelf-core-worker
template:
metadata:
labels:
app: airshelf-core-worker
spec:
imagePullSecrets:
- name: cr-pull-secret
containers:
- name: airshelf-core-worker
image: ${CI_REGISTRY_IMAGE}/airshelf-core-api:latest
imagePullPolicy: Always
# Celery worker connects to the external (Volcano managed) Redis broker
# configured via the airshelf-core-env secret. Uses `args` (not `command`)
# so the image ENTRYPOINT still runs but skips migrate/collectstatic ($1=celery).
args: ["celery", "-A", "airshelf.celery:app", "worker", "-l", "info", "--concurrency", "4"]
envFrom:
- secretRef:
name: airshelf-core-env
livenessProbe:
exec:
command: ["sh", "-c", "celery -A airshelf.celery:app inspect ping"]
initialDelaySeconds: 40
periodSeconds: 60
timeoutSeconds: 15
failureThreshold: 3
resources:
requests:
memory: "256Mi"
cpu: "100m"
limits:
memory: "768Mi"
cpu: "1000m"