Redirecting command output in docker

I want to do some simple logging for my server which is a small Flask app running in a Docker container.

Here is the Dockerfile

# Dockerfile
FROM dreen/flask
MAINTAINER dreen
WORKDIR /srv

# Get source
RUN mkdir -p /srv
COPY perfektimprezy.tar.gz /srv/perfektimprezy.tar.gz
RUN tar x -f perfektimprezy.tar.gz
RUN rm perfektimprezy.tar.gz

# Run server
EXPOSE 80
CMD ["python", "index.py", "1>server.log", "2>server.log"]

As you can see on the last line I redirect stderr and stdout to a file. Now I run this container and shell into it

docker run -d -p 80:80 perfektimprezy
docker exec -it "... id of container ..." bash

And observe the following things:

The server is running and the website working

There is no /srv/server.log

ps aux | grep python yields:

root         1  1.6  3.2  54172 16240 ?        Ss   13:43   0:00 python index.py 1>server.log 2>server.log
root        12  1.9  3.3 130388 16740 ?        Sl   13:43   0:00 /usr/bin/python index.py 1>server.log 2>server.log
root        32  0.0  0.0   8860   388 ?        R+   13:43   0:00 grep --color=auto python

But there are no logs... HOWEVER, if I docker attach to the container I can see the app generating output in the console.

How do I properly redirect stdout/err to a file when using Docker?


I'd like to question your use-case. Why do you need to redirect stderr/stdout to a log file in the container? Merged stderr and stdout are available outside the container using "docker logs", and they can be redirected to a file (outside of the container.) More details can be found here


Here's some info on a related use case: You want to redirect the output inside the container, which is running on Docker Cloud.

In my case, I'm executing a long-running data analysis script (Java) on Docker Cloud. Each container is supposed to write its own results file. I'm using the java:8-jre image and overwrite the 'Run command' with the following line:

bash -c "java -jar /code/analyzer.jar > /data/results-$RANDOM.txt"

By using bash's $RANDOM variable, I can scale up the number of containers in Docker Cloud as much as I want, while still collecting the results individually.