MarketAlly.AIPlugin.Extensions/MarketAlly.AIPlugin.Context/kubernetes/deployment.yaml

324 lines
7.6 KiB
YAML
Executable File

apiVersion: apps/v1
kind: Deployment
metadata:
name: context-plugin
namespace: marketally
labels:
app: context-plugin
version: v1.0.0
component: ai-plugin
spec:
replicas: 3
strategy:
type: RollingUpdate
rollingUpdate:
maxUnavailable: 1
maxSurge: 1
selector:
matchLabels:
app: context-plugin
template:
metadata:
labels:
app: context-plugin
version: v1.0.0
annotations:
prometheus.io/scrape: "true"
prometheus.io/port: "8081"
prometheus.io/path: "/metrics"
spec:
serviceAccountName: context-plugin
securityContext:
runAsNonRoot: true
runAsUser: 1000
runAsGroup: 1000
fsGroup: 1000
containers:
- name: context-plugin
image: ghcr.io/marketally/context-plugin:latest
imagePullPolicy: Always
ports:
- name: http
containerPort: 8080
protocol: TCP
- name: metrics
containerPort: 8081
protocol: TCP
env:
- name: ASPNETCORE_ENVIRONMENT
value: "Production"
- name: CONTEXT_STORAGE_PATH
value: "/app/data/.context"
- name: CONTEXT_LOG_LEVEL
valueFrom:
configMapKeyRef:
name: context-plugin-config
key: log-level
- name: CONTEXT_MAX_CONTEXT_SIZE
valueFrom:
configMapKeyRef:
name: context-plugin-config
key: max-context-size
- name: CONTEXT_RETENTION_DAYS
valueFrom:
configMapKeyRef:
name: context-plugin-config
key: retention-days
- name: CONTEXT_ENABLE_ENCRYPTION
valueFrom:
configMapKeyRef:
name: context-plugin-config
key: enable-encryption
- name: CONTEXT_ENCRYPTION_KEY
valueFrom:
secretKeyRef:
name: context-plugin-secrets
key: encryption-key
- name: OPENAI_API_KEY
valueFrom:
secretKeyRef:
name: context-plugin-secrets
key: openai-api-key
optional: true
resources:
requests:
memory: "256Mi"
cpu: "100m"
limits:
memory: "512Mi"
cpu: "500m"
volumeMounts:
- name: context-data
mountPath: /app/data
- name: context-logs
mountPath: /app/logs
- name: config
mountPath: /app/config
readOnly: true
livenessProbe:
httpGet:
path: /health
port: metrics
initialDelaySeconds: 60
periodSeconds: 30
timeoutSeconds: 10
failureThreshold: 3
readinessProbe:
httpGet:
path: /ready
port: metrics
initialDelaySeconds: 10
periodSeconds: 10
timeoutSeconds: 5
failureThreshold: 3
startupProbe:
httpGet:
path: /health
port: metrics
initialDelaySeconds: 30
periodSeconds: 10
timeoutSeconds: 10
failureThreshold: 6
volumes:
- name: context-data
persistentVolumeClaim:
claimName: context-plugin-data
- name: context-logs
emptyDir: {}
- name: config
configMap:
name: context-plugin-config
imagePullSecrets:
- name: ghcr-secret
affinity:
podAntiAffinity:
preferredDuringSchedulingIgnoredDuringExecution:
- weight: 100
podAffinityTerm:
labelSelector:
matchExpressions:
- key: app
operator: In
values:
- context-plugin
topologyKey: kubernetes.io/hostname
---
apiVersion: v1
kind: Service
metadata:
name: context-plugin-service
namespace: marketally
labels:
app: context-plugin
annotations:
prometheus.io/scrape: "true"
prometheus.io/port: "8081"
spec:
type: ClusterIP
ports:
- port: 8080
targetPort: 8080
protocol: TCP
name: http
- port: 8081
targetPort: 8081
protocol: TCP
name: metrics
selector:
app: context-plugin
---
apiVersion: v1
kind: ConfigMap
metadata:
name: context-plugin-config
namespace: marketally
data:
log-level: "Information"
max-context-size: "50000"
retention-days: "90"
enable-encryption: "true"
enable-caching: "true"
cache-expiration-minutes: "30"
max-concurrent-operations: "10"
enable-compression: "true"
enable-sensitive-data-detection: "true"
auto-encrypt-sensitive-data: "true"
enable-semantic-search: "false"
enable-fuzzy-matching: "true"
fuzzy-matching-threshold: "0.7"
---
apiVersion: v1
kind: Secret
metadata:
name: context-plugin-secrets
namespace: marketally
type: Opaque
data:
# Base64 encoded values - replace with actual values
encryption-key: bXktc2VjcmV0LWVuY3J5cHRpb24ta2V5 # my-secret-encryption-key
openai-api-key: "" # Optional: Base64 encoded OpenAI API key
---
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
name: context-plugin-data
namespace: marketally
spec:
accessModes:
- ReadWriteOnce
resources:
requests:
storage: 10Gi
storageClassName: standard
---
apiVersion: v1
kind: ServiceAccount
metadata:
name: context-plugin
namespace: marketally
labels:
app: context-plugin
---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRole
metadata:
name: context-plugin
rules:
- apiGroups: [""]
resources: ["pods", "nodes", "services"]
verbs: ["get", "list", "watch"]
- apiGroups: ["apps"]
resources: ["deployments", "replicasets"]
verbs: ["get", "list", "watch"]
---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRoleBinding
metadata:
name: context-plugin
roleRef:
apiGroup: rbac.authorization.k8s.io
kind: ClusterRole
name: context-plugin
subjects:
- kind: ServiceAccount
name: context-plugin
namespace: marketally
---
apiVersion: networking.k8s.io/v1
kind: Ingress
metadata:
name: context-plugin-ingress
namespace: marketally
annotations:
kubernetes.io/ingress.class: nginx
cert-manager.io/cluster-issuer: letsencrypt-prod
nginx.ingress.kubernetes.io/ssl-redirect: "true"
nginx.ingress.kubernetes.io/force-ssl-redirect: "true"
nginx.ingress.kubernetes.io/rate-limit: "100"
nginx.ingress.kubernetes.io/rate-limit-window: "1m"
spec:
tls:
- hosts:
- context-api.marketally.com
secretName: context-plugin-tls
rules:
- host: context-api.marketally.com
http:
paths:
- path: /
pathType: Prefix
backend:
service:
name: context-plugin-service
port:
number: 8080
---
apiVersion: policy/v1
kind: PodDisruptionBudget
metadata:
name: context-plugin-pdb
namespace: marketally
spec:
minAvailable: 2
selector:
matchLabels:
app: context-plugin
---
apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata:
name: context-plugin-hpa
namespace: marketally
spec:
scaleTargetRef:
apiVersion: apps/v1
kind: Deployment
name: context-plugin
minReplicas: 3
maxReplicas: 10
metrics:
- type: Resource
resource:
name: cpu
target:
type: Utilization
averageUtilization: 70
- type: Resource
resource:
name: memory
target:
type: Utilization
averageUtilization: 80
behavior:
scaleDown:
stabilizationWindowSeconds: 300
policies:
- type: Percent
value: 10
periodSeconds: 60
scaleUp:
stabilizationWindowSeconds: 60
policies:
- type: Percent
value: 50
periodSeconds: 60