在 Google App Engine 上部署 Flask 和 Tensorflow 服务

Question

I want to deploy Flask API with gunicorn and tensorflow serving to Google App Engine (Flex). I wrote a Dockerfile and startup.sh but fails to deploy. I increased memory to 6GB and set timeout 2 min for gunicorn, but it doesn't help.

Dockerfile runs successfully but startup.sh doesn't launch both of gunicorn and tensorflow serving. Can anybody point out what's wrong in sartup.sh?

Dockerfile

FROM gcr.io/google-appengine/python

RUN virtualenv /env -p python3.7
ENV VIRTUAL_ENV /env
ENV PATH /env/bin:$PATH

RUN apt-get update
ADD app/ /app/
RUN pip install --upgrade pip
RUN pip install -r /app/requirements.txt

RUN echo "deb http://storage.googleapis.com/tensorflow-serving-apt stable tensorflow-model-server tensorflow-model-server-universal" | tee /etc/apt/sources.list.d/tensorflow-serving.list && \
curl https://storage.googleapis.com/tensorflow-serving-apt/tensorflow-serving.release.pub.gpg | apt-key add -
RUN apt-get update
RUN apt-get install -y tensorflow-model-server

RUN apt-get install -y nginx
COPY app/default /etc/nginx/sites-available/default

WORKDIR /root

RUN chmod -R a+r /var/www/html

COPY startup.sh /startup.sh
RUN chmod 744 /startup.sh
RUN cd /
CMD /startup.sh

startup.sh

#!/bin/bash
/etc/init.d/nginx start
cd /app && nohup gunicorn app:app --bind 127.0.0.1:8081 --workers 1 --timeout 120 &
nohup tensorflow_model_server \
  --rest_api_port=8501 \
  --model_name=bird_net \
  --model_base_path=/app/saved_model &

Standard error output

$ gcloud app deploy
...
(ommited)
...
bc9f3e1065bb: Pushed
latest: digest: sha256:8dd67b5292199744a51d58c3cafb5ff17b87c4b39e35589c0d44e646dc1dd272 size: 5567
DONE
---------------------------------------------------------------------------------------------------------------------------------------------------------------

Updating service [default] (this may take several minutes)...failed.                                                                                          
ERROR: (gcloud.app.deploy) Error Response: [9] 
Application startup error! Code: APP_CONTAINER_CRASHED
 * Starting nginx nginx
   ...done.

Answer 1

I solved. Docker needs at least one foreground process.

#!/bin/bash
set -m
/etc/init.d/nginx start
cd /app
gunicorn app:app --bind 127.0.0.1:8081 --workers 1 --timeout 120 &
tensorflow_model_server \
  --rest_api_port=8501 \
  --model_name=birdnet \
  --model_base_path=/app/saved_model
fg %1

References
https://docs.docker.com/config/containers/multi-service_container/