112 lines
3.2 KiB
YAML
112 lines
3.2 KiB
YAML
# services/openclaw/ollama-deployment.yaml
|
|
apiVersion: apps/v1
|
|
kind: Deployment
|
|
metadata:
|
|
name: openclaw-ollama
|
|
namespace: openclaw
|
|
labels:
|
|
app: openclaw-ollama
|
|
spec:
|
|
replicas: 1
|
|
revisionHistoryLimit: 2
|
|
strategy:
|
|
type: Recreate
|
|
selector:
|
|
matchLabels:
|
|
app: openclaw-ollama
|
|
template:
|
|
metadata:
|
|
labels:
|
|
app: openclaw-ollama
|
|
annotations:
|
|
ai.bstein.dev/model: qwen2.5:1.5b-instruct-q4_0
|
|
ai.bstein.dev/gpu: Jetson pool (titan-20/21)
|
|
spec:
|
|
runtimeClassName: nvidia
|
|
affinity:
|
|
nodeAffinity:
|
|
requiredDuringSchedulingIgnoredDuringExecution:
|
|
nodeSelectorTerms:
|
|
- matchExpressions:
|
|
- key: kubernetes.io/hostname
|
|
operator: In
|
|
values:
|
|
- titan-20
|
|
- titan-21
|
|
volumes:
|
|
- name: models
|
|
emptyDir: {}
|
|
initContainers:
|
|
- name: warm-model
|
|
image: ollama/ollama@sha256:2c9595c555fd70a28363489ac03bd5bf9e7c5bdf2890373c3a830ffd7252ce6d
|
|
imagePullPolicy: IfNotPresent
|
|
env:
|
|
- name: OLLAMA_HOST
|
|
value: 0.0.0.0
|
|
- name: OLLAMA_MODELS
|
|
value: /root/.ollama
|
|
- name: OLLAMA_MODEL
|
|
value: qwen2.5:1.5b-instruct-q4_0
|
|
- name: NVIDIA_VISIBLE_DEVICES
|
|
value: all
|
|
- name: NVIDIA_DRIVER_CAPABILITIES
|
|
value: compute,utility
|
|
command:
|
|
- /bin/sh
|
|
- -c
|
|
- |
|
|
set -e
|
|
ollama serve >/tmp/ollama.log 2>&1 &
|
|
sleep 6
|
|
ollama pull "${OLLAMA_MODEL}"
|
|
pkill ollama || true
|
|
volumeMounts:
|
|
- name: models
|
|
mountPath: /root/.ollama
|
|
resources:
|
|
requests:
|
|
cpu: "1"
|
|
memory: 4Gi
|
|
nvidia.com/gpu.shared: 1
|
|
limits:
|
|
cpu: "4"
|
|
memory: 10Gi
|
|
nvidia.com/gpu.shared: 1
|
|
containers:
|
|
- name: ollama
|
|
image: ollama/ollama@sha256:2c9595c555fd70a28363489ac03bd5bf9e7c5bdf2890373c3a830ffd7252ce6d
|
|
imagePullPolicy: IfNotPresent
|
|
ports:
|
|
- name: http
|
|
containerPort: 11434
|
|
env:
|
|
- name: OLLAMA_HOST
|
|
value: 0.0.0.0
|
|
- name: OLLAMA_KEEP_ALIVE
|
|
value: 6h
|
|
- name: OLLAMA_MODELS
|
|
value: /root/.ollama
|
|
- name: NVIDIA_VISIBLE_DEVICES
|
|
value: all
|
|
- name: NVIDIA_DRIVER_CAPABILITIES
|
|
value: compute,utility
|
|
volumeMounts:
|
|
- name: models
|
|
mountPath: /root/.ollama
|
|
readinessProbe:
|
|
httpGet:
|
|
path: /api/tags
|
|
port: 11434
|
|
initialDelaySeconds: 15
|
|
periodSeconds: 10
|
|
timeoutSeconds: 5
|
|
resources:
|
|
requests:
|
|
cpu: "2"
|
|
memory: 8Gi
|
|
nvidia.com/gpu.shared: 1
|
|
limits:
|
|
cpu: "6"
|
|
memory: 12Gi
|
|
nvidia.com/gpu.shared: 1
|