Toggle navigation

Installation & Configuration

Getting Started

Superset is tested against Python 2.7 and Python 3.4. Airbnb currently uses 2.7.* in production. We do not plan on supporting Python 2.6.


Superset is designed to be highly available. It is “cloud-native” as it has been designed scale out in large, distributed environments, and works well inside containers. While you can easily test drive Superset on a modest setup or simply on your laptop, there’s virtually no limit around scaling out the platform. Superset is also cloud-native in the sense that it is flexible and lets you choose your web server (Gunicorn, Nginx, Apache), your metadata database engine (MySQL, Postgres, MariaDB, …), your message queue (Redis, RabbitMQ, SQS, …), your results backend (S3, Redis, Memcached, …), your caching layer (memcached, Redis, …), works well with services like NewRelic, StatsD and DataDog, and has the ability to run analytic workloads against most popular database technologies.

Superset is battle tested in large environments with hundreds of concurrent users. Airbnb’s production environment runs inside Kubernetes and serves 600+ daily active users viewing over 100K charts a day.

The Superset web server and the Superset Celery workers (optional) are stateless, so you can scale out by running on as many servers as needed.

OS dependencies

Superset stores database connection information in its metadata database. For that purpose, we use the cryptography Python library to encrypt connection passwords. Unfortunately this library has OS level dependencies.

You may want to attempt the next step (“Superset installation and initialization”) and come back to this step if you encounter an error.

Here’s how to install them:

For Debian and Ubuntu, the following command will ensure that the required dependencies are installed:

sudo apt-get install build-essential libssl-dev libffi-dev python-dev python-pip libsasl2-dev libldap2-dev

Ubuntu 16.04 If you have python3.5 installed alongside with python2.7, as is default on Ubuntu 16.04 LTS, run this command also

sudo apt-get install build-essential libssl-dev libffi-dev python3.5-dev python-pip libsasl2-dev libldap2-dev

otherwhise build for cryptography fails.

For Fedora and RHEL-derivatives, the following command will ensure that the required dependencies are installed:

sudo yum upgrade python-setuptools
sudo yum install gcc gcc-c++ libffi-devel python-devel python-pip python-wheel openssl-devel libsasl2-devel openldap-devel

OSX, system python is not recommended. brew’s python also ships with pip

brew install pkg-config libffi openssl python
env LDFLAGS="-L$(brew --prefix openssl)/lib" CFLAGS="-I$(brew --prefix openssl)/include" pip install cryptography==1.9

Windows isn’t officially supported at this point, but if you want to attempt it, download, and run python which may need admin access. Then run the following:

C:\> pip install cryptography

# You may also have to create C:\Temp
C:\> md C:\Temp

Python virtualenv

It is recommended to install Superset inside a virtualenv. Python 3 already ships virtualenv, for Python 2 you need to install it. If it’s packaged for your operating systems install it from there otherwise you can install from pip:

pip install virtualenv

You can create and activate a virtualenv by:

# virtualenv is shipped in Python 3 as pyvenv
virtualenv venv
. ./venv/bin/activate

On windows the syntax for activating it is a bit different:


Once you activated your virtualenv everything you are doing is confined inside the virtualenv. To exit a virtualenv just type deactivate.

Python’s setup tools and pip

Put all the chances on your side by getting the very latest pip and setuptools libraries.:

pip install --upgrade setuptools pip

Superset installation and initialization

Follow these few simple steps to install Superset.:

# Install superset
pip install superset

# Create an admin user (you will be prompted to set username, first and last name before setting a password)
fabmanager create-admin --app superset

# Initialize the database
superset db upgrade

# Load some data to play with
superset load_examples

# Create default roles and permissions
superset init

# Start the web server on port 8088, use -p to bind to another port
superset runserver

# To start a development web server, use the -d switch
# superset runserver -d

After installation, you should be able to point your browser to the right hostname:port http://localhost:8088, login using the credential you entered while creating the admin account, and navigate to [UNKNOWN NODE title_reference]. This action should bring in all of your datasources for Superset to be aware of, and they should show up in [UNKNOWN NODE title_reference], from where you can start playing with your data!

A proper WSGI HTTP Server

While you can setup Superset to run on Nginx or Apache, many use Gunicorn, preferably in async mode, which allows for impressive concurrency even and is fairly easy to install and configure. Please refer to the documentation of your preferred technology to set up this Flask WSGI application in a way that works well in your environment.

While the [UNKNOWN NODE title_reference] command act as an quick wrapper around [UNKNOWN NODE title_reference], it doesn’t expose all the options you may need, so you’ll want to craft your own [UNKNOWN NODE title_reference] command in your production environment. Here’s an async setup known to work well:

gunicorn \
              -w 10 \
              -k gevent \
              --timeout 120 \
              -b \
              --limit-request-line 0 \
              --limit-request-field_size 0 \
              --statsd-host localhost:8125 \

Refer to the [Gunicorn documentation]( for more information.

Note that gunicorn does not work on Windows so the [UNKNOWN NODE title_reference] command is not expected to work in that context. Also note that the development web server ([UNKNOWN NODE title_reference]) is not intended for production use.

If not using gunicorn, you may want to disable the use of flask-compress by setting [UNKNOWN NODE title_reference] in your [UNKNOWN NODE title_reference]

Flask-AppBuilder Permissions

By default every time the Flask-AppBuilder (FAB) app is initialized the permissions and views are added automatically to the backend and associated with the ‘Admin’ role. The issue however is when you are running multiple concurrent workers this creates a lot of contention and race conditions when defining permissions and views.

To alleviate this issue, the automatic updating of permissions can be disabled by setting the SUPERSET_UPDATE_PERMS environment variable to [UNKNOWN NODE title_reference]. The value [UNKNOWN NODE title_reference] enables it, [UNKNOWN NODE title_reference] disables it. Note if undefined the functionality is enabled to maintain backwards compatibility.

In a production environment initialization could take on the following form:

export SUPERSET_UPDATE_PERMS=1 superset init

export SUPERSET_UPDATE_PERMS=0 gunicorn -w 10 … superset:app

Configuration behind a load balancer

If you are running superset behind a load balancer or reverse proxy (e.g. NGINX or ELB on AWS), you may need to utilise a healthcheck endpoint so that your load balancer knows if your superset instance is running. This is provided at /health which will return a 200 response containing “OK” if the webserver is running.

If the load balancer is inserting X-Forwarded-For/X-Forwarded-Proto headers, you should set [UNKNOWN NODE title_reference] in the superset config file to extract and use the headers.

In case that the reverse proxy is used for providing ssl encryption, an explicit definition of the [UNKNOWN NODE title_reference] may be required. For the Apache webserver this can be set as follows:

 RequestHeader set X-Forwarded-Proto “https”


To configure your application, you need to create a file (module) and make sure it is in your PYTHONPATH. Here are some of the parameters you can copy / paste in that configuration module:

# Superset specific config
ROW_LIMIT = 5000


# Flask App Builder configuration
# Your App secret key
SECRET_KEY = '\2\1thisismyscretkey\1\2\e\y\y\h'

# The SQLAlchemy connection string to your database backend
# This connection defines the path to the database that stores your
# superset metadata (slices, connections, tables, dashboards, ...).
# Note that the connection information to connect to the datasources
# you want to explore are managed directly in the web UI
SQLALCHEMY_DATABASE_URI = 'sqlite:////path/to/superset.db'

# Flask-WTF flag for CSRF
# Add endpoints that need to be exempt from CSRF protection

# Set this API key to enable Mapbox visualizations

This file also allows you to define configuration parameters used by Flask App Builder, the web framework used by Superset. Please consult the Flask App Builder Documentation for more information on how to configure Superset.

Please make sure to change:

  • SQLALCHEMY_DATABASE_URI, by default it is stored at ~/.superset/superset.db
  • SECRET_KEY, to a long random string

In case you need to exempt endpoints from CSRF, e.g. you are running a custom auth postback endpoint, you can add them to WTF_CSRF_EXEMPT_LIST


Database dependencies

Superset does not ship bundled with connectivity to databases, except for Sqlite, which is part of the Python standard library. You’ll need to install the required packages for the database you want to use as your metadata database as well as the packages needed to connect to the databases you want to access through Superset.

Here’s a list of some of the recommended packages.

databasepypi packageSQLAlchemy URI prefix
MySQLpip install mysqlclientmysql://
Postgrespip install psycopg2postgresql+psycopg2://
Prestopip install pyhivepresto://
Oraclepip install cx_Oracleoracle://
Redshiftpip install sqlalchemy-redshiftpostgresql+psycopg2://
MSSQLpip install pymssqlmssql://
Impalapip install impylaimpala://
SparkSQLpip install pyhivejdbc+hive://
Greenplumpip install psycopg2postgresql+psycopg2://
Athenapip install "PyAthenaJDBC>1.0.9"awsathena+jdbc://
Verticapip install sqlalchemy-vertica-pythonvertica+vertica_python://
ClickHousepip install sqlalchemy-clickhouseclickhouse://
Kylinpip install kylinpykylin://

Note that many other database are supported, the main criteria being the existence of a functional SqlAlchemy dialect and Python driver. Googling the keyword sqlalchemy in addition of a keyword that describes the database you want to connect to should get you to the right place.

(AWS) Athena

The connection string for Athena looks like this


Where you need to escape/encode at least the s3_staging_dir, i.e.,

s3://... -> s3%3A//...


Superset uses Flask-Cache for caching purpose. Configuring your caching backend is as easy as providing a CACHE_CONFIG, constant in your that complies with the Flask-Cache specifications.

Flask-Cache supports multiple caching backends (Redis, Memcached, SimpleCache (in-memory), or the local filesystem). If you are going to use Memcached please use the [UNKNOWN NODE title_reference] client library as [UNKNOWN NODE title_reference] does not handle storing binary data correctly. If you use Redis, please install the redis Python package:

pip install redis

For setting your timeouts, this is done in the Superset metadata and goes up the “timeout searchpath”, from your slice configuration, to your data source’s configuration, to your database’s and ultimately falls back into your global default defined in CACHE_CONFIG.

        'CACHE_TYPE': 'redis',
        'CACHE_DEFAULT_TIMEOUT': 60 * 60 * 24, # 1 day default (in secs)
        'CACHE_KEY_PREFIX': 'superset_results',
        'CACHE_REDIS_URL': 'redis://localhost:6379/0',

Deeper SQLAlchemy integration

It is possible to tweak the database connection information using the parameters exposed by SQLAlchemy. In the Database edit view, you will find an extra field as a JSON blob.

This JSON string contains extra configuration elements. The engine_params object gets unpacked into the sqlalchemy.create_engine call, while the metadata_params get unpacked into the sqlalchemy.MetaData call. Refer to the SQLAlchemy docs for more information.

Schemas (Postgres & Redshift)

Postgres and Redshift, as well as other database, use the concept of schema as a logical entity on top of the database. For Superset to connect to a specific schema, there’s a schema parameter you can set in the table form.

External Password store for SQLAlchemy connections

It is possible to use an external store for you database passwords. This is useful if you a running a custom secret distribution framework and do not wish to store secrets in Superset’s meta database.

Example: Write a function that takes a single argument of type sqla.engine.url and returns the password for the given connection string. Then set SQLALCHEMY_CUSTOM_PASSWORD_STORE in your config file to point to that function.

def example_lookup_password(url):
    secret = <<get password from external framework>>
    return 'secret'

SQLALCHEMY_CUSTOM_PASSWORD_STORE = example_lookup_password

SSL Access to databases

This example worked with a MySQL database that requires SSL. The configuration may differ with other backends. This is what was put in the extra parameter

    "metadata_params": {},
    "engine_params": {
              "sslrootcert": "/path/to/my/pem"


  • From the UI, enter the information about your clusters in the [UNKNOWN NODE title_reference] menu by hitting the + sign.
  • Once the Druid cluster connection information is entered, hit the [UNKNOWN NODE title_reference] menu item to populate
  • Navigate to your datasources

Note that you can run the superset refresh_druid command to refresh the metadata from your Druid cluster(s)


The extra CORS Dependency must be installed:


The following keys in [UNKNOWN NODE title_reference] can be specified to configure CORS:

  • ENABLE_CORS: Must be set to True in order to enable CORS
  • CORS_OPTIONS: options passed to Flask-CORS ([UNKNOWN NODE title_reference])


Superset allows you to add your own middleware. To add your own middleware, update the ADDITIONAL_MIDDLEWARE key in your [UNKNOWN NODE title_reference]. ADDITIONAL_MIDDLEWARE should be a list of your additional middleware classes.

For example, to use AUTH_REMOTE_USER from behind a proxy server like nginx, you have to add a simple middleware class to add the value of HTTP_X_PROXY_REMOTE_USER (or any other custom header from the proxy) to Gunicorn’s REMOTE_USER environment variable:

class RemoteUserMiddleware(object):
    def __init__(self, app): = app
    def __call__(self, environ, start_response):
        user = environ.pop('HTTP_X_PROXY_REMOTE_USER', None)
        environ['REMOTE_USER'] = user
        return, start_response)

ADDITIONAL_MIDDLEWARE = [RemoteUserMiddleware, ]

Adapted from


Upgrading should be as straightforward as running:

pip install superset --upgrade
superset db upgrade
superset init


SQL Lab is a powerful SQL IDE that works with all SQLAlchemy compatible databases. By default, queries are executed in the scope of a web request so they may eventually timeout as queries exceed the maximum duration of a web request in your environment, whether it’d be a reverse proxy or the Superset server itself.

On large analytic databases, it’s common to run queries that execute for minutes or hours. To enable support for long running queries that execute beyond the typical web request’s timeout (30-60 seconds), it is necessary to configure an asynchronous backend for Superset which consist of:

  • one or many Superset worker (which is implemented as a Celery worker), and can be started with the superset worker command, run superset worker --help to view the related options
  • a celery broker (message queue) for which we recommend using Redis or RabbitMQ
  • a results backend that defines where the worker will persist the query results

Configuring Celery requires defining a CELERY_CONFIG in your Both the worker and web server processes should have the same configuration.

class CeleryConfig(object):
    BROKER_URL = 'redis://localhost:6379/0'
    CELERY_IMPORTS = ('superset.sql_lab', )
    CELERY_RESULT_BACKEND = 'redis://localhost:6379/0'
    CELERY_ANNOTATIONS = {'tasks.add': {'rate_limit': '10/s'}}

CELERY_CONFIG = CeleryConfig

To setup a result backend, you need to pass an instance of a derivative of werkzeug.contrib.cache.BaseCache to the RESULTS_BACKEND configuration key in your It’s possible to use Memcached, Redis, S3 (, memory or the file system (in a single server-type setup or for testing), or to write your own caching interface. Your may look something like:

# On S3
from s3cache.s3cache import S3Cache
S3_CACHE_BUCKET = 'foobar-superset'
S3_CACHE_KEY_PREFIX = 'sql_lab_result'

# On Redis
from werkzeug.contrib.cache import RedisCache
    host='localhost', port=6379, key_prefix='superset_results')

Note that it’s important that all the worker nodes and web servers in the Superset cluster share a common metadata database. This means that SQLite will not work in this context since it has limited support for concurrency and typically lives on the local file system.

Also note that SQL Lab supports Jinja templating in queries, and that it’s possible to overload the default Jinja context in your environment by defining the JINJA_CONTEXT_ADDONS in your superset configuration. Objects referenced in this dictionary are made available for users to use in their SQL.

    'my_crazy_macro': lambda x: x*2,

Making your own build

For more advanced users, you may want to build Superset from sources. That would be the case if you fork the project to add features specific to your environment.:

# assuming $SUPERSET_HOME as the root of the repo
cd $SUPERSET_HOME/superset/assets
yarn run build
python install


Blueprints are Flask’s reusable apps. Superset allows you to specify an array of Blueprints in your superset_config module. Here’s an example on how this can work with a simple Blueprint. By doing so, you can expect Superset to serve a page that says “OK” at the /simple_page url. This can allow you to run other things such as custom data visualization applications alongside Superset, on the same server.

from flask import Blueprint
simple_page = Blueprint('simple_page', __name__,
@simple_page.route('/', defaults={'page': 'index'})
def show(page):
    return "Ok"

BLUEPRINTS = [simple_page]

StatsD logging

Superset is instrumented to log events to StatsD if desired. Most endpoints hit are logged as well as key events like query start and end in SQL Lab.

To setup StatsD logging, it’s a matter of configuring the logger in your

from superset.stats_logger import StatsdStatsLogger
STATS_LOGGER = StatsdStatsLogger(host='localhost', port=8125, prefix='superset')

Note that it’s also possible to implement you own logger by deriving superset.stats_logger.BaseStatsLogger.