本文章只记录和官网有差别的部分
官网为1.6.0版本,不包含fate-serving部分:
为了使用KubeFATE部署FATE,我们需要以下工具:
- MiniKube: v1.7.3
- kubectl: v1.17.3
- kubefate:
- 发布版本: v1.6.1
- 服务版本: v1.4.2
- 命令行版本: v1.4.2
首先去https://github.com/FederatedAI/KubeFATE/releases下载
修改kubefate.yaml中的版本1.4.1为1.4.2(应该是个bug,不该的话* kubefate service version=v1.4.1 就是老版本
)
- image: federatedai/kubefate:v1.4.1
cluster-9999.yaml
name: fate-9999
namespace: fate-9999
chartName: fate
chartVersion: v1.6.1
partyId: 9999
registry: "hub.c.163.com/federatedai"
imageTag: "1.6.1-release"
pullPolicy:
#imagePullSecrets:
#- name: myregistrykey
persistence: false
istio:
enabled: false
podSecurityPolicy:
enabled: false
modules:
- rollsite
- clustermanager
- nodemanager
- mysql
- python
- fateboard
- client
backend: eggroll
# host:
# fateboard: 9999.fateboard.kubefate.net
# client: 9999.notebook.kubefate.net
rollsite:
type: NodePort
nodePort: 30091
partyList:
- partyId: 10000
partyIp: 192.168.6.72
partyPort: 30101
python:
type: NodePort
httpNodePort: 30097
grpcNodePort: 30092
servingIp: 192.168.6.72
servingPort: 30095
cluster-10000.yaml
name: fate-10000
namespace: fate-10000
chartName: fate
chartVersion: v1.6.1
partyId: 10000
registry: "hub.c.163.com/federatedai"
imageTag: "1.6.1-release"
pullPolicy:
#imagePullSecrets:
#- name: myregistrykey
persistence: false
istio:
enabled: false
podSecurityPolicy:
enabled: false
modules:
- rollsite
- clustermanager
- nodemanager
- mysql
- python
- fateboard
- client
backend: eggroll
# host:
# fateboard: 10000.fateboard.kubefate.net
# client: 10000.notebook.kubefate.net
rollsite:
type: NodePort
nodePort: 30101
partyList:
- partyId: 9999
partyIp: 192.168.6.72
partyPort: 30091
python:
type: NodePort
httpNodePort: 30107
grpcNodePort: 30102
servingIp: 192.168.6.72
servingPort: 30105
cluster-serving-9999.yaml
name: fate-serving-9999
namespace: fate-serving-9999
chartName: fate-serving
chartVersion: v2.0.4
partyId: 9999
registry: "hub.c.163.com/federatedai"
imageTag:
pullPolicy:
#imagePullSecrets:
#- name: myregistrykey
persistence: false
istio:
enabled: false
podSecurityPolicy:
enabled: false
modules:
- servingProxy
- servingRedis
- servingServer
- servingZookeeper
- servingAdmin
servingAdmin:
ingressHost: 9999.serving-admin.kubefate.net
username: admin
password: admin
servingProxy:
nodePort: 30096
type: NodePort
ingerssHost: 9999.serving-proxy.kubefate.net
partyList:
- partyId: 10000
partyIp: 192.168.6.72
partyPort: 30106
nodeSelector: {}
servingServer:
type: NodePort
nodePort: 30095
fateflow:
ip: 192.168.6.72
port: 30097
subPath: ""
existingClaim: ""
storageClass: "serving-server"
accessMode: ReadWriteOnce
size: 1Gi
nodeSelector: {}
servingRedis:
password: fate_dev
nodeSelector: {}
subPath: ""
existingClaim: ""
storageClass: "serving-redis"
accessMode: ReadWriteOnce
size: 1Gi
servingZookeeper:
subPath: ""
existingClaim: ""
storageClass: "serving-zookeeper"
accessMode: ReadWriteOnce
size: 1Gi
cluster-serving-10000.yaml
name: fate-serving-10000
namespace: fate-serving-10000
chartName: fate-serving
chartVersion: v2.0.4
partyId: 10000
registry: "hub.c.163.com/federatedai"
imageTag:
pullPolicy:
#imagePullSecrets:
#- name: myregistrykey
persistence: false
istio:
enabled: false
podSecurityPolicy:
enabled: false
modules:
- servingProxy
- servingRedis
- servingServer
- servingZookeeper
- servingAdmin
servingAdmin:
ingressHost: 10000.serving-admin.kubefate.net
username: admin
password: admin
servingProxy:
nodePort: 30106
type: NodePort
ingerssHost: 10000.serving-proxy.kubefate.net
partyList:
- partyId: 9999
partyIp: 192.168.6.72
partyPort: 30096
nodeSelector: {}
servingServer:
type: NodePort
nodePort: 30105
fateflow:
ip: 192.168.6.72
port: 30107
subPath: ""
existingClaim: ""
storageClass: "serving-server"
accessMode: ReadWriteOnce
size: 1Gi
nodeSelector: {}
servingRedis:
password: fate_dev
nodeSelector: {}
subPath: ""
existingClaim: ""
storageClass: "serving-redis"
accessMode: ReadWriteOnce
size: 1Gi
servingZookeeper:
subPath: ""
existingClaim: ""
storageClass: "serving-zookeeper"
accessMode: ReadWriteOnce
size: 1Gi
安装集群
kubefate cluster install -f ./cluster-9999.yaml
kubefate cluster install -f ./cluster-10000.yaml
kubefate cluster install -f ./cluster-serving-9999.yaml
kubefate cluster install -f ./cluster-serving-10000.yaml
安装好后
kubectl get po --all-namespaces