analytics/HOSTING.md
Chandra Tungathurthi 517d4db262
Selfhosted version improvements (#89)
* first commit with test and compile job

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* adding 'prepare' stage

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* updated ci script to include "test" compile phase

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* adding environment variables for connecting to postgresql

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* updated ci config for postgres

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* using non-alpine version of elixir

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* re-using the 'compile' artifacts and added explict env variables for testing

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* removing redundant deps fetching from common code

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* formatting using mix.format -- beware no-code changes!

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* added release config

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* adding consistent env variable for Database

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* more cleaning up of environment variables

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* Adding releases config for enabling releases

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* cleaning up env configs

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* Cleaned up config and prepared config for releases

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* updated CI script with new config for test

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* Added Dockerfile for creating production docker image

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* Adding "docker" build job yay!

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* using non-slim version of debian and installing webpack

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* Adding overlays for migrations on releases

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* restricting the docker built to master branch only

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* typo fix

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* adding "Hosting.md" to explain hosting instructions

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* removed the default comments

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* Added documentation related to env variables

* updated documentation and fixed typo

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* updated documentation

* Bumping up elixir version as `overlays` are only supported in latest version

read release notes: https://github.com/elixir-lang/elixir/releases/tag/v1.10.0

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* Adding tarball assembly during release

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* updated HOSTING.md

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* Added support for db migration

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* minor corrections

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* initializing admin user

Admin user has been added in the "migration" phase. A default user is automatically created in the process. One can provide the related env variables, else a new one will be automatically created for you.

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* Initial base domain update - phase#1

These changes are only meant for correct operating it under self-hosting. There are many other cosmetic changes, that require updates to email, site and other places where the original website and author is used.

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* Using dedicated config variable `base_domain` instead

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* adding base_domain to releases config

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* removing the dedicated config "base_domain", relying on endpoint host

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* Removed the usage of "Mix" in code!

It is bad practice to use "mix" module inside the code as in actual release this module is unavailable. Replacing this with a config environment variable

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* Added support for SMTP via Bamboo Smtp Adapter

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* Capturing SMTP errors via Sentry

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* Minor updates

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* Adding junit formatter -- useful for generating test reports

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* adding documentation for default user

* Resolve "Gitlab Adoption: Add supported services in "Security & Compliance""

* bumping up the debian version to fix issues

fixing some vulnerabilities identified by the scanning tools

* More updates for self-hosting

Changes in most of the places to suit self-hosting. Although, there are some which have been left-off.

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* quick-dirty-fix!

* bumping up the db connect timeout

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* bumping up the db connect timeout

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* bumping up the db connect timeout

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* bumping up timeout - skipping MRs :-/

* removing restrictions on watching for changes

this stuff isn't working

* Update HOSTING.md

* renamed the module name

* reverting formatting-whitespace changes

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* reverting the name to release

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* adding docker-compose.yml and related instructions

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* using `plausible_url` instead of assuming `https`

this is because, it is much to test in local dev machines and in most cases there's already a layer above which is capable for `https` termination and http -> https upgrade

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* WIP: merging changes from upstream

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* wip: more changes

* Pushing in changes from upstream

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* changes to ci for testing

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* cleaning up and finishing clickhouse integration

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* updating readme with hosting details

* removing deleted files from upstream

* minor config adjustments

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* formatting changes

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* changing the connection strategy for clickhouse during release

since clickhouse integration doesn't have an ecto support, we need to prepare the db _before_ the clickhouse migration. One workaround is to connect to a default db on init and then create a db

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* formatting

* cleanup and added separated migration to setup

* Big improvements to selfhosting

- added ability for disabling
  - authentication completely
  - registration
  - landing page

- formatting cleanups

* Big improvements to selfhosting

- added ability for disabling
  - authentication completely
  - registration
  - landing page

- formatting cleanups

* changing smtp auth  to optional

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* removed stale templates and permanently removed landing page

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>

* removed stale templates and permanently removed landing page

Signed-off-by: Chandra Tungathurthi <tckb@tgrthi.me>
2020-07-02 11:21:11 +03:00

9.6 KiB

status
Beta

Plausible Analytics

Self-hosting is possible based on the docker images and are automatically pushed into Dockerhub registry for all commits on master branch. At the moment, latest is the only tag on DockerHub as we haven't reached a stable release of self-hosted Plausible yet.

Architecture

Plausible runs as a single server, backed by two databases: PostgreSQL for user data and ClickhouseDB for the stats. When you download and run the docker image you also need to provide connection details for these two databases.

Most hosting providers will offer a managed PostgreSQL instance, but it's not as simple with Clickhouse. You can install Clickhouse on a VPS, run it using their official Docker image, or use a managed service provided by Yandex Cloud. Aiven has announced that they are planning offer a managed ClickHouse service in the future and more hosting providers are following suit.

As of June 2020, here's the setup of the official cloud version of Plausible for reference:

  • Web server: Digital Ocean Droplet w/ 1vCPU & 2GB RAM. Managed via the official Docker app.
  • User database: Digital Ocean Managed PostgreSQL w/ 1vCPU & 1GB RAM.
  • Stats database: Digital Ocean Droplet w/ 6vCPU & 16GB RAM. Installed on Ubuntu 18.04 using the official tutorial

Total cost: $105/mo

Building Docker image

Besides the DockerHub registry, one can build docker image from Dockerfile.

Up and Running

The repo supplies with a Docker Compose file and the sample environment variables , this serves as a sample for running Plausible with Docker.

  • Running the setup takes care of the initial migration steps, this needs to be executed only once, on the first run.

    docker-compose run --rm setup
    docker-compose down
    
  • After the setup, you can start plausible as --

    docker-compose up -d plausible 
    

    after a successful startup (can take upto 5 mins), plausible is available at port 80, navigate to http://localhost.

  • stopping plausible --

    docker-compose down
    
  • purging and removing everything --

    docker-compose down
    docker volume rm plausible_event-data -f
    docker volume rm plausible_db-data -f
    

Note:

  • #1 you need to stop plausible and restart plausible if you change the environment variables.
  • #2 With docker-compose, you need to remove the existing container and rebuild if you want your changes need to be reflected:
    docker rmi -f  plausible_plausible:latest
    docker-compose up -d plausible 
    

Non-docker building

It is possible to create a release artifact by running a release.

MIX_ENV=prod mix release plausible

the release will create the pre-packed artifact at _build/prod/rel/plausible/bin/plausible, the release will also create a tarball at _build/prod/ for convenience.

Note, that you have to feed in the related environment variables (see below Environment Variables)

Database Migration

On the initial setup, a migration step is necessary to create database and table schemas needed for initial bootup. Normally, this done by mix aliases like ecto.setup defined in the mix.exs. As this not available in "released" artifact, plausible_migration.ex facilitates this process. The overlay scripts take care of these.

After the release, these are available under _build/prod/rel/plausible --

_build/prod/rel/plausible/createdb.sh
_build/prod/rel/plausible/migrate.sh
_build/prod/rel/plausible/rollback.sh
_build/prod/rel/plausible/seed.sh

the same is available in the docker images as follows --

docker run plausible:master-12add db createdb
docker run plausible:master-12add db migrate
docker run plausible:master-12add db rollback
docker run plausible:master-12add db seed

Environment Variables

Plausible relies on the several services for operating, the expected environment variables are explaiend below.

Server

Following are the variables that can be used to configure the availability of the server.

  • HOST (String)
    • The hosting address of the server. For running on local system, this can be set to localhost. In production systems, this can be your ingress host.
  • SCHEME (String)
    • The scheme of the URL, either http or https. When using a reverse proxy with https, it'll be required to set this. defaults to http
  • PORT (Number)
    • The port on which the server is available.
  • SECRET_KEY_BASE (String)
  • ENVIRONMENT (String)
    • The current running environment. defaults to prod
  • APP_VERSION (String)
    • The version of the app running. defaults to current docker tag
  • DISABLE_AUTH (Boolean String)
    • Disables authentication completely, no registration, login will be shown. defaults to false
    • Note: This option is not recommended for production deployments.
  • DISABLE_REGISTRATION
    • Disables registration of new users, keep your admin credentials handy ;) defaults to false

Default User Generation

For self-hosting, a default user is generated during the [Database Migration](#Database Migration) to access Plausible. To be noted that, a default user is a user whose trial period expires in 100 Years ;). It is highly recommended that you configure these parameters.

  • ADMIN_USER_NAME
    • The default ("admin") username. if not provided, one will be generated for you
  • ADMIN_USER_EMAIL
    • The default ("admin") user email. if not provided, one will be generated for you
  • ADMIN_USER_PWD
    • The default ("admin") user password. if not provided, one will be generated for you

Mailer/SMTP Setup

  • MAILER_ADAPTER (String)
    • The adapter used for sending out e-mails. Available: Bamboo.PostmarkAdapter / Bamboo.SMTPAdapter
  • MAILER_EMAIL (String)
    • The email id to use for as from address of all communications from Plausible.

In case of Bamboo.SMTPAdapter you need to supply the following variables:

  • SMTP_HOST_ADDR (String)
    • The host address of your smtp server.
  • SMTP_HOST_PORT (Number)
    • The port of your smtp server.
  • SMTP_USER_NAME (String)
    • The username/email for smtp auth.
  • SMTP_USER_PWD (String)
    • The password for smtp auth.
  • SMTP_HOST_SSL_ENABLED (Boolean String)
    • If ssl is enabled for connecting to Smtp, defaults to false
  • SMTP_RETRIES (Number)
    • Number of retries to make until mailer gives up. defaults to 2
  • SMTP_MX_LOOKUPS_ENABLED (Boolean String)
    • If MX lookups should be done before sending out emails. defaults to false

Database

Plausible uses postgresql as database for storing all the user-data. Use the following the variables to configure it.

  • DATABASE_URL (String)
    • The repo Url as dictated here
  • DATABASE_POOL_SIZE (Number)
    • A default pool size for connecting to the database, defaults to 10, a higher number is recommended for a production system.
  • DATABASE_TLS_ENABLED (Boolean String)
    • A flag that says whether to connect to the database via TLS, read here

For performance reasons, all the analytics events are stored in clickhouse:

  • CLICKHOUSE_DATABASE_HOST (String)
  • CLICKHOUSE_DATABASE_NAME (String)
  • CLICKHOUSE_DATABASE_USER (String)
  • CLICKHOUSE_DATABASE_PASSWORD (String)
  • CLICKHOUSE_DATABASE_POOLSIZE (Number)
    • A default pool size for connecting to the database, defaults to 10, a higher number is recommended for a production system.

IP Geolocation

Plausible uses the GeoLite2 database created by MaxMind for enriching analytics data with visitor countries. Their end-user license does not make it very easy to just package the database along with an open-source product. This is why, if you want to get country data for your analytics, you need to create an account and download their GeoLite2 Country database.

Once you have the database, mount it on the Plausible docker image and configure the path of the database file:

  • GEOLITE2_COUNTRY_DB (String)

If the Geolite database is not configured, no country data will be captured.

External Services

  • Google Client
    • GOOGLE_CLIENT_ID
    • GOOGLE_CLIENT_SECRET
  • Sentry
    • SENTRY_DSN
  • Paddle
    • PADDLE_VENDOR_AUTH_CODE
  • PostMark, only in case of Bamboo.PostmarkAdapter mail adapter.
    • POSTMARK_API_KEY

Apart from these, there are also the following integrations

  • Twitter
    • TWITTER_CONSUMER_KEY
    • TWITTER_CONSUMER_SECRET
    • TWITTER_ACCESS_TOKEN
    • TWITTER_ACCESS_TOKEN_SECRET
  • Slack
    • SLACK_WEBHOOK