69 lines
2.3 KiB
YAML
69 lines
2.3 KiB
YAML
# services/ai-llm/batch-deployment.yaml
|
|
apiVersion: apps/v1
|
|
kind: Deployment
|
|
metadata:
|
|
name: ollama-batch
|
|
namespace: ai
|
|
spec:
|
|
replicas: 1
|
|
revisionHistoryLimit: 2
|
|
strategy:
|
|
type: Recreate
|
|
selector:
|
|
matchLabels: {app: ollama-batch}
|
|
template:
|
|
metadata:
|
|
labels: {app: ollama-batch}
|
|
spec:
|
|
automountServiceAccountToken: false
|
|
nodeSelector:
|
|
kubernetes.io/hostname: titan-23
|
|
tolerations:
|
|
- {key: veles.bstein.dev/simulation, operator: Equal, value: "true", effect: NoSchedule}
|
|
securityContext:
|
|
seccompProfile: {type: RuntimeDefault}
|
|
containers:
|
|
- name: ollama
|
|
image: ollama/ollama@sha256:0c0a83210471fb50226bcdc2d6611d20ab13ae87e024cc304c94a6a5765c5e65
|
|
command: [/bin/bash, -ec]
|
|
args:
|
|
- |
|
|
while ! test -f /models/.pilot-v1-ready; do sleep 5; done
|
|
exec ollama serve
|
|
env:
|
|
- {name: OLLAMA_HOST, value: "0.0.0.0:11434"}
|
|
- {name: OLLAMA_MODELS, value: /models}
|
|
- {name: OLLAMA_NO_CLOUD, value: "1"}
|
|
- {name: OLLAMA_CONTEXT_LENGTH, value: "32768"}
|
|
- {name: OLLAMA_MAX_LOADED_MODELS, value: "1"}
|
|
- {name: OLLAMA_NUM_PARALLEL, value: "1"}
|
|
- {name: OLLAMA_MAX_QUEUE, value: "1"}
|
|
- {name: OLLAMA_KEEP_ALIVE, value: "10m"}
|
|
- {name: OLLAMA_LOAD_TIMEOUT, value: "10m"}
|
|
- {name: OLLAMA_DEBUG, value: "false"}
|
|
ports:
|
|
- {name: http, containerPort: 11434}
|
|
readinessProbe:
|
|
httpGet: {path: /api/version, port: http}
|
|
periodSeconds: 10
|
|
timeoutSeconds: 3
|
|
securityContext:
|
|
allowPrivilegeEscalation: false
|
|
readOnlyRootFilesystem: true
|
|
capabilities: {drop: [ALL]}
|
|
resources:
|
|
requests: {cpu: "16", memory: 48Gi}
|
|
limits: {cpu: "16", memory: 48Gi}
|
|
volumeMounts:
|
|
- {name: models, mountPath: /models, readOnly: true}
|
|
- {name: tmp, mountPath: /tmp}
|
|
- {name: identity, mountPath: /root/.ollama}
|
|
volumes:
|
|
- name: models
|
|
persistentVolumeClaim:
|
|
claimName: ollama-batch-titan23
|
|
- name: tmp
|
|
emptyDir: {sizeLimit: 1Gi}
|
|
- name: identity
|
|
emptyDir: {medium: Memory, sizeLimit: 1Mi}
|