To learn more about why Alpine images are discouraged for Python read the note at the end: 🚨 Alpine Python Warning.
These tags are no longer supported:
Note: There are tags for each build date. If you need to "pin" the Docker image version you use, you can select one of those tags. E.g.
Docker image with uWSGI and Nginx for web applications in Python (as Flask) in a single container. Optionally with Alpine Linux.
The combination of uWSGI with Nginx is a common way to deploy Python web applications like Flask and Django. It is widely used in the industry and would give you decent performance. (*)
There is also an Alpine version. If you want it, check the tags from above.
This image was created to be the base image for tiangolo/uwsgi-nginx-flask but could be used as the base image for any other (WSGI-based) Python web application, like Django.
If you are starting a new project, you might benefit from a newer and faster framework based on ASGI instead of WSGI (Flask and Django are WSGI-based).
You could use an ASGI framework like:
FastAPI, or Starlette, would give you about 800% (8x) the performance achievable with this image (tiangolo/uwsgi-nginx). You can see the third-party benchmarks here.
Also, if you want to use new technologies like WebSockets it would be easier (and possible) with a newer framework based on ASGI, like FastAPI or Starlette. As the standard ASGI was designed to be able to handle asynchronous code like the one needed for WebSockets.
If you need to use an older WSGI-based framework like Flask or Django (instead of something based on ASGI) and you need to have the best performance possible, you can use the alternative image: tiangolo/meinheld-gunicorn.
tiangolo/meinheld-gunicorn will give you about 400% (4x) the performance of this image.
Docker Hub image: https://hub.docker.com/r/tiangolo/uwsgi-nginx/
You are probably using Kubernetes or similar tools. In that case, you probably don't need this image (or any other similar base image). You are probably better off building a Docker image from scratch.
If you have a cluster of machines with Kubernetes, Docker Swarm Mode, Nomad, or other similar complex system to manage distributed containers on multiple machines, then you will probably want to handle replication at the cluster level instead of using a process manager in each container that starts multiple worker processes, which is what this Docker image does.
In those cases (e.g. using Kubernetes) you would probably want to build a Docker image from scratch, installing your dependencies, and running a single process instead of this image.
For example, using Gunicorn you could have a file
# Gunicorn config variables loglevel = "info" errorlog = "-" # stderr accesslog = "-" # stdout worker_tmp_dir = "/dev/shm" graceful_timeout = 120 timeout = 120 keepalive = 5 threads = 3
And then you could have a
FROM python:3.9 WORKDIR /code COPY ./requirements.txt /code/requirements.txt RUN pip install --no-cache-dir --upgrade -r /code/requirements.txt COPY ./app /code/app CMD ["gunicorn", "--conf", "app/gunicorn_conf.py", "--bind", "0.0.0.0:80", "app.main:app"]
You can read more about these ideas in the FastAPI documentation about: FastAPI in Containers - Docker as the same ideas would apply to other web applications in containers.
You could want a process manager running multiple worker processes in the container if your application is simple enough that you don't need (at least not yet) to fine-tune the number of processes too much, and you can just use an automated default, and you are running it on a single server, not a cluster.
You could be deploying to a single server (not a cluster) with Docker Compose, so you wouldn't have an easy way to manage replication of containers (with Docker Compose) while preserving the shared network and load balancing.
Then you could want to have a single container with a process manager starting several worker processes inside, as this Docker image does.
You could also have other reasons that would make it easier to have a single container with multiple processes instead of having multiple containers with a single process in each of them.
For example (depending on your setup) you could have some tool like a Prometheus exporter in the same container that should have access to each of the requests that come.
In this case, if you had multiple containers, by default, when Prometheus came to read the metrics, it would get the ones for a single container each time (for the container that handled that particular request), instead of getting the accumulated metrics for all the replicated containers.
Then, in that case, it could be simpler to have one container with multiple processes, and a local tool (e.g. a Prometheus exporter) on the same container collecting Prometheus metrics for all the internal processes and exposing those metrics on that single container.
Read more about it all in the FastAPI documentation about: FastAPI in Containers - Docker, as the same concepts apply to other web applications in containers.
You don't have to clone this repo.
You can use this image as a base image for other images.
Assuming you have a file
requirements.txt, you could have a
Dockerfile like this:
FROM tiangolo/uwsgi-nginx:python3.9 COPY ./requirements.txt /app/requirements.txt RUN pip install --no-cache-dir --upgrade -r /app/requirements.txt COPY ./app /app # Your Dockerfile code...
By default it will try to find a uWSGI config file in
uwsgi.ini file will make it try to run a Python file in
If you are building a Flask web application you should use instead tiangolo/uwsgi-nginx-flask.
If you need to use a directory for your app different than
/app, you can override the uWSGI config file path with an environment variable
UWSGI_INI, and put your custom
uwsgi.ini file there.
For example, if you needed to have your application directory in
/application instead of
Dockerfile would look like:
FROM tiangolo/uwsgi-nginx:python3.9 ENV UWSGI_INI /application/uwsgi.ini COPY ./application /application WORKDIR /appapplication
uwsgi.ini file in
./application/uwsgi.ini would contain:
Note: it's important to include the
WORKDIR option, otherwise uWSGI will start the application in
By default, the image starts with 2 uWSGI processes running. When the server is experiencing a high load, it creates up to 16 uWSGI processes to handle it on demand.
If you need to configure these numbers you can use environment variables.
The starting number of uWSGI processes is controlled by the variable
UWSGI_CHEAPER, by default set to
The maximum number of uWSGI processes is controlled by the variable
UWSGI_PROCESSES, by default set to
Have in mind that
UWSGI_CHEAPER must be lower than
So, if, for example, you need to start with 4 processes and grow to a maximum of 64, your
Dockerfile could look like:
FROM tiangolo/uwsgi-nginx:python3.9 ENV UWSGI_CHEAPER 4 ENV UWSGI_PROCESSES 64 COPY ./app /app
In this image, Nginx is configured to allow unlimited upload file sizes. This is done because by default a simple Python server would allow that, so that's the simplest behavior a developer would expect.
If you need to restrict the maximum upload size in Nginx, you can add an environment variable
NGINX_MAX_UPLOAD and assign a value corresponding to the standard Nginx config
For example, if you wanted to set the maximum upload file size to 1 MB (the default in a normal Nginx installation), you would need to set the
NGINX_MAX_UPLOAD environment variable to the value
1m. Then the image would take care of adding the corresponding configuration file (this is done by the
Dockerfile would look something like:
FROM tiangolo/uwsgi-nginx:python3.9 ENV NGINX_MAX_UPLOAD 1m COPY ./app /app
By default, the container made from this image will listen on port 80.
To change this behavior, set the
LISTEN_PORT environment variable.
You might also need to create the respective
EXPOSE Docker instruction.
You can do that in your
Dockerfile, it would look something like:
FROM tiangolo/uwsgi-nginx:python3.9 ENV LISTEN_PORT 8080 EXPOSE 8080 COPY ./app /app
If you need to run anything before starting the app, you can add a file
prestart.sh to the directory
/app. The image will automatically detect and run it before starting everything.
For example, if you want to add database migrations that are run on startup (e.g. with Alembic, or Django migrations), before starting the app, you could create a
./app/prestart.sh file in your code directory (that will be copied by your
#! /usr/bin/env bash # Let the DB start sleep 10; # Run migrations alembic upgrade head
and it would wait 10 seconds to give the database some time to start and then run that
alembic command (you could update that to run Django migrations or any other tool you need).
If you need to run a Python script before starting the app, you could make the
/app/prestart.sh file run your Python script, with something like:
#! /usr/bin/env bash # Run custom Python script before starting python /app/my_custom_prestart_script.py
Note: The image uses
. to run the script (as in
. /app/prestart.sh), so for example, environment variables would persist. If you don't understand the previous sentence, you probably don't need it.
By default, Nginx will start one "worker process".
If you want to set a different number of Nginx worker processes you can use the environment variable
You can use a specific single number, e.g.:
ENV NGINX_WORKER_PROCESSES 2
or you can set it to the keyword
auto and it will try to autodetect the number of CPUs available and use that for the number of workers.
For example, using
auto, your Dockerfile could look like:
FROM tiangolo/uwsgi-nginx:python3.9 ENV NGINX_WORKER_PROCESSES auto COPY ./app /app
By default, Nginx will start with a maximum limit of 1024 connections per worker.
If you want to set a different number you can use the environment variable
ENV NGINX_WORKER_CONNECTIONS 2048
It cannot exceed the current limit on the maximum number of open files. See how to configure it in the next section.
The number connections per Nginx worker cannot exceed the limit on the maximum number of open files.
You can change the limit of open files with the environment variable
ENV NGINX_WORKER_OPEN_FILES 2048
If you need to configure Nginx further, you can add
*.conf files to
/etc/nginx/conf.d/ in your
Just have in mind that the default configurations are created during startup in a file at
/etc/nginx/conf.d/upload.conf. So you shouldn't overwrite them. You should name your
*.conf file with something different than
upload.conf, for example:
Note: if you are customizing Nginx, maybe copying configurations from a blog or a StackOverflow answer, have in mind that you probably need to use the configurations specific to uWSGI, instead of those for other modules, like for example,
If you need to configure Nginx even further, completely overriding the defaults, you can add a custom Nginx configuration to
It will be copied to
/etc/nginx/nginx.conf and used instead of the generated one.
Have in mind that, in that case, this image won't generate any of the Nginx configurations, it will only copy and use your configuration file.
That means that all the environment variables described above that are specific to Nginx won't be used.
It also means that it won't use additional configurations from files in
/etc/nginx/conf.d/*.conf, unless you explicitly have a section in your custom file
If you want to add a custom
/app/nginx.conf file but don't know where to start from, you can use the
nginx.conf used for the tests and customize it or modify it further.
The combination of uWSGI with Nginx is a common way to deploy Python web applications.
Nginx is a web server, it takes care of the HTTP connections and also can serve static files directly and more efficiently.
uWSGI is an application server, that's what runs your Python code and it talks with Nginx.
Your Python code has the actual web application, and is run by uWSGI.
This image takes advantage of already slim and optimized existing Docker images (based on Debian as recommended by Docker) and implements Docker best practices.
It uses the official Python Docker image, installs uWSGI and on top of that, with the least amount of modifications, adds the official Nginx image (as of 2016-02-14).
And it controls all these processes with Supervisord.
There's the rule of thumb that you should have "one process per container".
That helps, for example, isolating an app and its database in different containers.
But if you want to have a "micro-services" approach you may want to have more than one process in one container if they are all related to the same "service", and you may want to include your Flask code, uWSGI and Nginx in the same container (and maybe run another container with your database).
That's the approach taken in this image.
This image has a default sample "Hello World" app in the container's
/app directory using the example in the uWSGI documentation.
You probably want to override it or delete it in your project.
It is there in case you run this image by itself and not as a base image for your own
Dockerfile, so that you get a sample app without errors.
In short: You probably shouldn't use Alpine for Python projects, instead use the
slim Docker image versions.
Do you want more details? Continue reading 👇
Alpine is more useful for other languages where you build a static binary in one Docker image stage (using multi-stage Docker building) and then copy it to a simple Alpine image, and then just execute that binary. For example, using Go.
But for Python, as Alpine doesn't use the standard tooling used for building Python extensions, when installing packages, in many cases Python (
pip) won't find a precompiled installable package (a "wheel") for Alpine. And after debugging lots of strange errors you will realize that you have to install a lot of extra tooling and build a lot of dependencies just to use some of these common Python packages. 😩
This means that, although the original Alpine image might have been small, you end up with a an image with a size comparable to the size you would have gotten if you had just used a standard Python image (based on Debian), or in some cases even larger. 🤯
And in all those cases, it will take much longer to build, consuming much more resources, building dependencies for longer, and also increasing its carbon footprint, as you are using more CPU time and energy for each build. 🌳
If you want slim Python images, you should instead try and use the
slim versions that are still based on Debian, but are smaller. 🤓
All the image tags, configurations, environment variables and application options are tested.
Updates are announced in the releases.
You can click the "watch" button at the top right and select "Releases only" to receive an email notification when there's a new release.
1.17.10, based on latest Debian, Buster. PR #82.
tiangolo/uwsgi-nginx:python3.7-2019-09-28. PR #65.
Upgrade Travis. PR #60.
/app/prestart.shscript to run arbitrary code before starting the app (for example, Alembic - SQLAlchemy migrations). The documentation for the
/app/prestart.shis in the main README. PR #59.
2018-06-22: You can now use
NGINX_WORKER_CONNECTIONS to set the maximum number of Nginx worker connections and
NGINX_WORKER_OPEN_FILES to set the maximum number of open files. Thanks to ronlut in this PR.
2018-06-22: Make uWSGI require an app to run, instead of going in "full dynamic mode" while there was an error. Supervisord doesn't terminate itself but tries to restart uWSGI and shows the errors. Uses
need-app as suggested by luckydonald in this comment.
2018-01-14: There are now two Alpine based versions,
2017-08-09: You can set a custom maximum upload file size using an environment variable
NGINX_MAX_UPLOAD, by default it has a value of
0, that allows unlimited upload file sizes. This differs from Nginx's default value of 1 MB. It's configured this way because that's the simplest experience a developer that is not expert in Nginx would expect.
2017-08-09: Now you can override where to look for the
uwsgi.ini file, and with that, change the default directory from
/app to something else, using the envirnoment variable
2017-08-08: Supervisord now terminates uWSGI on
SIGTERM, so if you run
docker stop or something similar, it will actually stop everything, instead of waiting for Docker's timeout to kill the container.
2016-10-01: Now you can override default
uwsgi.ini parameters from the file in
2016-08-16: There's now an image tag for Python 3.5, based on the official image for Python 3.5. So now you can use this image for your projects in Python 2.7 and Python 3.5.
2016-08-16: Use dynamic a number of worker processes for uWSGI, from 2 to 16 depending on load. This should work for most cases. This helps especially when there are some responses that are slow and take some time to be generated, this change allows all the other responses to keep fast (in a new process) without having to wait for the first (slow) one to finish.
Also, it now uses a base
uwsgi.ini file under
/etc/uwsgi/ with most of the general configurations, so, the
/app (the one you could need to modify) is now a lot simpler.
2016-04-05: Nginx and uWSGI logs are now redirected to stdout, allowing to use
This project is licensed under the terms of the Apache license.