Paul Payne
9fd046c378
New version slots (major version jumps): - gitea: 1.24.3 -> 28.2.0 (slot 1 -> 28) - ghost: 5.130.6 -> 6.72.0 (slot 5 -> 6) - formbricks: 3.6.0 -> 6.1.0 (slot 3 -> 6) - mattermost: 10.8.1 -> 12.0.1 (slot 10 -> 12) - zulip: 9.4 -> 12.3 (slot 9 -> 12) - immich: 1.135.3 -> 3.3.5 (slot 1 -> 3) - openproject: 16.1.1 -> 17.9.3 (slot 16 -> 17) - listmonk: 5.0.3 -> 6.1.0 (slot 5 -> 6) - discourse: 3.5.3 -> 2026.10.0 (slot 3 -> 2026, calver) Same-slot large bumps: - vllm: 0.9.2 -> 0.32.0 - longhorn: v1.8.1 -> v1.13.0 Infrastructure: - cert-manager: v1.17.2 -> v1.21.2 - coredns: v1.12.0 -> v1.14.7 - externaldns: v0.13.4 -> v0.23.0 - metallb: v0.15.0 -> v0.16.1 - mysql: 9.1.0 -> 9.7.4 - node-feature-discovery: v0.17.3 -> v0.20.1 - nvidia-device-plugin: v0.17.1 -> v0.20.1
93 lines
2.6 KiB
YAML
93 lines
2.6 KiB
YAML
apiVersion: apps/v1
|
|
kind: Deployment
|
|
metadata:
|
|
name: vllm
|
|
spec:
|
|
replicas: 1
|
|
strategy:
|
|
type: Recreate
|
|
selector:
|
|
matchLabels:
|
|
component: inference
|
|
template:
|
|
metadata:
|
|
labels:
|
|
component: inference
|
|
spec:
|
|
runtimeClassName: nvidia
|
|
enableServiceLinks: false
|
|
securityContext:
|
|
seccompProfile:
|
|
type: RuntimeDefault
|
|
nodeSelector:
|
|
feature.node.kubernetes.io/pci-0300_10de.present: "true"
|
|
containers:
|
|
- name: vllm
|
|
image: vllm/vllm-openai:v0.32.0
|
|
imagePullPolicy: IfNotPresent
|
|
securityContext:
|
|
allowPrivilegeEscalation: false
|
|
capabilities:
|
|
drop:
|
|
- ALL
|
|
args:
|
|
- --model={{ .model }}
|
|
- --max-model-len={{ .maxModelLen }}
|
|
- --tensor-parallel-size=1
|
|
- --gpu-memory-utilization={{ .gpuMemoryUtilization }}
|
|
- --enforce-eager
|
|
- --reasoning-parser=deepseek_r1
|
|
- --api-key=$(VLLM_API_KEY)
|
|
env:
|
|
- name: VLLM_API_KEY
|
|
valueFrom:
|
|
secretKeyRef:
|
|
name: vllm-secrets
|
|
key: apiKey
|
|
- name: VLLM_TORCH_DTYPE
|
|
value: "auto"
|
|
- name: VLLM_WORKER_CONCURRENCY
|
|
value: "1"
|
|
- name: HF_HOME
|
|
value: "/root/.cache/huggingface"
|
|
- name: NUMBA_CACHE_DIR
|
|
value: "/tmp"
|
|
volumeMounts:
|
|
- name: hf-cache
|
|
mountPath: /root/.cache/huggingface
|
|
ports:
|
|
- name: http
|
|
containerPort: 8000
|
|
resources:
|
|
requests:
|
|
cpu: "{{ .cpuRequest }}"
|
|
memory: "{{ .memoryRequest }}"
|
|
nvidia.com/gpu: {{ .gpuCount }}
|
|
limits:
|
|
cpu: "{{ .cpuLimit }}"
|
|
memory: "{{ .memoryLimit }}"
|
|
nvidia.com/gpu: {{ .gpuCount }}
|
|
readinessProbe:
|
|
httpGet:
|
|
path: /health
|
|
port: http
|
|
initialDelaySeconds: 300
|
|
periodSeconds: 30
|
|
timeoutSeconds: 10
|
|
failureThreshold: 5
|
|
livenessProbe:
|
|
httpGet:
|
|
path: /health
|
|
port: http
|
|
initialDelaySeconds: 1800
|
|
periodSeconds: 60
|
|
timeoutSeconds: 10
|
|
failureThreshold: 3
|
|
volumes:
|
|
- name: hf-cache
|
|
emptyDir: {}
|
|
tolerations:
|
|
- key: "nvidia.com/gpu"
|
|
operator: "Exists"
|
|
effect: "NoSchedule"
|