基于gunicorn+flask+docker模型高并发部署

此项目的是基于已有深度学习环境进行迁移，利用gunicorn+flask+docker的形式进行模型部署，实现模型的高并发。

代码结构如下：

├── Dockerfile
├── keras_MuLabelClass.tar.gz     -------服务器上打包好的环境
├── Miniconda3-py37_4.11.0-Linux-x86_64.sh ----------conda
├── multi-label-classification-event.tar    ----------镜像
├── multi-label-classification-event-type   ----------源码
├── multi-label-classification-event-type.tar.gz  ----------打包的源码
└── requirements.txt  

1、深度学习环境迁移

1.1、基于conda-pack算法环境打包

迁出机器安装打包工具

pip install conda-pack

安装好之后打包需要迁出的环境（-n 之后为 虚拟环境名字 -o 之后为打包出来的文件名）

conda pack -n keras_MuLabelClass -o keras_MuLabelClass.tar.gz

迁入机器部分——解压、部署环境

迁入mkdir /.../miniconda3/envs/keras_MuLabelClass
tar -xzf keras_MuLabelClass.tar.gz  
conda activate keras_MuLabelClass

1.2、基于linux命令的算法环境打包

直接把/xxxx/anaconda/envs 下的keras_MuLabelClass打包压缩即可
zip -r  pytorch_cpu.zip /xxxx/miniconda/envs/keras_MuLabelClass
将pyenv.zip 拷贝到另外一个系统上并解压到目录下/xxxx/anaconda/envs/
unzip -d /xxxx/miniconda/envs/keras_MuLabelClass.zip 

2、并发代码：gunicorn_conf.py

bind = '0.0.0.0:3114' #监听内网端口
backlog = 1024
workers = 3  ### 并行工作进程数
worker_class = 'sync'# 工作模式协程使用sync模式，还可以使用gevent 模式，默认的是sync模式
worker_connections = 1000 # 设置最大并发量
timeout = 121 #超时
keepalive = 2

spew = False

daemon = False # 设置守护进程,将进程交给supervisor管理
raw_env = [
    'DJANGO_SECRET_KEY=something',
    'SPAM=eggs',
]
pidfile = None
umask = 0
user = None
group = None
tmp_upload_dir = None


# 设置访问日志和错误信息日志路径
errorlog = '-'
loglevel = 'info'
accesslog = '-'
access_log_format = '%(h)s %(l)s %(u)s %(t)s "%(r)s" %(s)s %(b)s "%(f)s" "%(a)s"'

proc_name = None

def post_fork(server, worker):
    server.log.info("Worker spawned (pid: %s)", worker.pid)

def pre_fork(server, worker):
    pass

def pre_exec(server):
    server.log.info("Forked child, re-executing.")

def when_ready(server):
    server.log.info("Server is ready. Spawning workers")

def worker_int(worker):
    worker.log.info("worker received INT or QUIT signal")

    ## get traceback info
    import threading, sys, traceback
    id2name = {th.ident: th.name for th in threading.enumerate()}
    code = []
    for threadId, stack in sys._current_frames().items():
        code.append("\n# Thread: %s(%d)" % (id2name.get(threadId,""),
            threadId))
        for filename, lineno, name, line in traceback.extract_stack(stack):
            code.append('File: "%s", line %d, in %s' % (filename,
                lineno, name))
            if line:
                code.append("  %s" % (line.strip()))
    worker.log.debug("\n".join(code))

def worker_abort(worker):
    worker.log.info("worker received SIGABRT signal")

3、dockerfile构建

From centos:centos7
  
# miniconda
ADD Miniconda3-py37_4.11.0-Linux-x86_64.sh /opt/
RUN cd opt && bash Miniconda3-py37_4.11.0-Linux-x86_64.sh -b -u -p $HOME/miniconda


# keras_MuLabelClass env
RUN mkdir /root/miniconda/envs/keras_MuLabelClass
ADD keras_MuLabelClass.tar.gz /root/miniconda/envs/keras_MuLabelClass
RUN $HOME/miniconda/bin/conda init && sed -i '$a\conda activate keras_MuLabelClass' /root/.bashrc

ADD multi-label-classification-event-type.tar.gz /root/
RUN rm -rf /opt/*

WORKDIR /root/multi-label-classification-event-type/
RUN yum -y install vim
#/root/miniconda/envs/keras_MuLabelClass/lib/python3.7/site-packages/keras/engine

RUN rm -rf /root/miniconda/envs/keras_MuLabelClass/lib/python3.7/site-packages/keras/engine/saving.py

RUN mv saving.py /root/miniconda/envs/keras_MuLabelClass/lib/python3.7/site-packages/keras/engine
RUN chmod +x ApiSever.py
CMD ["gunicorn", "-c", "./gunicorn_conf.py", "ApiSever:app"]

EXPOSE 3114

4、项目部署

docker run -it -d --name njdmultilabel_v3 -p 3114:3114 --restart=always -v  ....../scene_classifi