基础

• KuberFate_v1.6.1
• docker-compose方式部署
• 训练服务及推理服务位于同一台服务器,docker容器内网互通;共部署四台;

编写zookeeper启动文件,配置文件

#zk的docker-compose.yml文件内容

version: '3'
networks:
  fate-network:
    external: true
    name: fate-network
services:
  zoo1:
    image: zookeeper
    restart: always
    container_name: zoo1
    expose:
      - 2181
    ports:
      - 2181:2181
      - 8000:2181
    networks:
      - fate-network
    volumes:
      - /data/projects/zookeeper/zoo1/data:/data
      - /data/projects/zookeeper/zoo1/datalog:/datalog
      - ./confs/zoo.cfg:/apache-zookeeper-3.7.0-bin/conf/zoo.cfg
    environment:
      ZOO_MY_ID: 1
      ZOO_SERVERS: server.1=zoo1:2888:3888;2181 server.2=zoo2:2888:3888;2181 server.3=zoo3:2888:3888;2181
  zoo2:
    image: zookeeper
    restart: always
    container_name: zoo2
    networks:
      - fate-network
    expose:
      - 2181
    ports:
      - 2182:2181
    volumes:
      - /data/projects/zookeeper/zoo2/data:/data
      - /data/projects/zookeeper/zoo2/datalog:/datalog
      - ./confs/zoo.cfg:/apache-zookeeper-3.7.0-bin/conf/zoo.cfg
    environment:
      ZOO_MY_ID: 2
      ZOO_SERVERS: server.1=zoo1:2888:3888;2181 server.2=zoo2:2888:3888;2181 server.3=zoo3:2888:3888;2181
  zoo3:
	image: zookeeper
    restart: always
    container_name: zoo3
    networks:
      - fate-network
    expose:
      - 2181
    ports:
      - 2183:2181
    volumes:
      - /data/projects/zookeeper/zoo3/data:/dada
      - /data/projects/zookeeper/zoo3/datalog:/datalog
      - ./confs/zoo.cfg:/apache-zookeeper-3.7.0-bin/conf/zoo.cfg
    environment:
      ZOO_MY_ID: 3
      ZOO_SERVERS: server.1=zoo1:2888:3888;2181 server.2=zoo2:2888:3888;2181 server.3=zoo3:2888:3888;2181

zk配置文件内容

# The number of milliseconds of each 
ticktickTime=2000
# The number of ticks that the initial 
# synchronization phase can take
initLimit=10
# The number of ticks that can pass between 
# sending a request and getting an acknowledgement
syncLimit=5
# the directory where the snapshot is stored.
# do not use /tmp for storage, /tmp here is just 
# example sakes.dataDir=/tmp/zookeeper
# the port at which the clients will connect
clientPort=2181
dataDir=/data/projects/zookeeper
# the maximum number of client connections.
# increase this if you need to handle more clients
#maxClientCnxns=60
#
# Be sure to read the maintenance section of the 
# administrator guide before turning on autopurge.
#
# http://zookeeper.apache.org/doc/current/zookeeperAdmin.html#sc_maintenance
#
# The number of snapshots to retain in dataDir
#autopurge.snapRetainCount=3
# Purge task interval in hours
# Set to "0" to disable auto purge feature
#autopurge.purgeInterval=1
admin.enableServer=false

目录结构如图:

使用docker-compose up -d 直接启动,并日志核实;

FATE serving-server及serving-proxy配置参数修改

serving-server配置文件:

#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS, 
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
port=8000
#serviceRoleName=serving
# cache
#remoteModelInferenceResultCacheSwitch=false
#cache.type=local
#model.cache.path=
# local cache
#local.cache.maxsize=10000
#local.cache.expire=30
#local.cache.interval=3
# external cache
redis.ip=redis
redis.port=6379
### configure this parameter to use cluster mode
#redis.cluster.nodes=127.0.0.1:6379,127.0.0.1:6380,127.0.0.1:6381,127.0.0.1:6382,127.0.0.1:6383,127.0.0.1:6384
### this password is common in stand-alone mode and cluster mode
redis.password=fate_dev
#redis.timeout=10
#redis.expire=3000
#redis.maxTotal=100
#redis.maxIdle=100
# external subsystem
#  proxy=serving-proxy:8879 #需要关闭掉
# adapter
feature.single.adaptor=com.webank.ai.fate.serving.adaptor.dataaccess.MockAdapter
feature.batch.adaptor=com.webank.ai.fate.serving.adaptor.dataaccess.MockBatchAdapter
# model transfer
model.transfer.url=http://172.21.16.15:9380/v1/model/transfer
# zk router
zk.url=zoo1:2181,zoo2:2181,zoo3:2181   #已修改
useRegister=true   #已修改
useZkRouter=true   #已修改
# zk acl
#acl.enable=false
#acl.username=
#acl.password=

serving-proxy配置文件修改

# limitations under the License.
#
# coordinator same as Party ID
coordinator=10000
server.port=8059
#inference.service.name=serving
#random, consistent
#routeType=random
#route.table=/data/projects/fate-serving/serving-proxy/conf/route_table.json
#auth.file=/data/projects/fate-serving/serving-proxy/conf/auth_config.json
# zk router
#useZkRouter=true
useRegister=true  # 修改
useZkRouter=true  # 修改
zk.url=zoo1:2181,zoo2:2181,zoo3:2181  # 修改
# zk acl
#acl.enable=false
#acl.username=
#acl.password=
# intra-partyid port
proxy.grpc.intra.port=8879  # 修改
# inter-partyid port
proxy.grpc.inter.port=8869   # 修改
# grpc
proxy.grpc.inference.timeout=3000   # 修改
proxy.grpc.inference.async.timeout=1000   # 修改
proxy.grpc.unaryCall.timeout=3000  # 修改
proxy.grpc.threadpool.coresize=50  # 修改
proxy.grpc.threadpool.maxsize=100  # 修改
proxy.grpc.threadpool.queuesize=10  # 修改
proxy.async.timeout=5000  # 修改
proxy.async.coresize=10  # 修改
proxy.async.maxsize=100  # 修改
proxy.grpc.batch.inference.timeout=10000  # 修改

编写serving-admin启动及配置文件

#serving-admin的docker-compose.yml 文件
version: '3'
networks:
  fate-network:
    external: true
    name: confs-10000_fate-network
services:
  serving-admin:
    image: "federatedai/serving-admin:2.0.4-release"
    volumes:
      - ./confs/serving-admin/conf/application.properties:/data/projects/fate/serving-admin/conf/application.properties
    ports:
      - "9300:8350"
    networks:
      - fate-network
    restart: always
    container_name: serving-admin

配置文件:

server.port=8350 
# cache
#local.cache.expire=300
# zk
#zk.url=zoo1:2181
zk.url=zoo1:2181,zoo2:2181,zoo3:2181
# zk acl
#acl.enable=false
#acl.username=
#acl.password=
# grpc
#grpc.timeout=5000
# username & password
admin.username=admin
admin.password=admin

使用docker-compose up -d 直接启动,并日志核实;注意,一定要在zk和FATE-SERVING启动成功后,最后再启动serving-admin服务;
访问ip:8350即可,配置文件中的账号密码登录;