OpenStreetMap Ansible playbook

This playbook is useful for setting up a complete, self-contained instance of OpenStreetMap, including:

the read/write API
the web interface
a tile rendering pipeline
PostgreSQL for data storage

Installing the stack manually takes a LOT of time, so hopefully this speeds up the process somewhat.

Structure

The playbook assumes a single machine is used for all components. It could probably be tweaked to move heavier components to their own machine if needed.

A quick overview:

The Apache web server sits in front of renderd and Mapnik. When tile requests hit the server, the mod_tile extension forwards them to renderd, which generates the tiles and returns them to the user. Generated tiles are cached on disk.
The data for the tiles is contained in a PostGIS/PostgreSQL database (the gis table), and can be imported from OSM Planet files (see below)
The stylesheet is built using openstreetmap-carto and is used when creating the tile images
Apache also serves the OpenStreetMap web application, which handles editing the map data.

Important: The web application and render pipeline actually use independent copies of map data (in incompatible formats). See below for how to persist edits from the application to the tile generator.

Usage

Requirements

You will need a server running Ubuntu 18.04. I used a VPS with 4GB of RAM and 2 vCPUs, which seems to work OK, although the more powerful the machine the faster tile generation will run.

It's probably a good idea to mount an external volume to store the PostgreSQL database. After importing OSM data for Oceania, my database was around ~40GB.

You will also need:

Ansible install locally
Python installed on the VPS (apt-get install python)

Running the playbook

Create a file called hosts in this directory, containing your server's IP address like:
```
[osm]
78.46.xxx.xxx
```
Run the playbook: ansible-playbook -i hosts site.yml

Importing initial data

If you want to prepopulate the server with OSM data, you will need a Planet file in PBF format.

You can download the full Planet (47GB) here or a regional extract here. The more heavily-mapped the area, the larger the extract will be.

Initially, import the Planet into the OSM application database (command from here).

osmosis --read-pbf <PLANET_FILE>.osm.pbf \
  --write-apidb host="localhost" database="openstreetmap" \
  user="root" password="osm12345" validateSchemaVersion="no"

If you've changed any of the defaults in defaults/main.yml, you will need to change them in the command above.

Updating tile data

After making edits to the OSM data, you will want to reflect that in the generated tiles. To convert OSM data to PostGIS:

# Export updated planet file
osmosis --read-apidb database="openstreetmap" user=root password="osm12345" validateSchemaVersion="no" \
        --write-pbf file="/mapserver/planet/planet.osm.pbf"

# Convert to PostGIS and load into database
osm2pgsql --create --database gis --slim --hstore --multi-geometry \
          --tag-transform-script /mapserver/styles/standard/openstreetmap-carto.lua \
          --style /mapserver/styles/standard/openstreetmap-carto.style "/mapserver/planet/planet.osm.pbf"

# Clear cache
rm -rf /mapserver/website/tiles/standard/*

(commands from here)

Improving performance

You can decrease tile rendering time by making some tweaks to PostgresSQL. See here.

Thanks

Thanks to user Ika-chan! from the OpenGeoFiction community, who documented most of the setup and usage process here. This playbook really just automates their hard work and brings it up to date with changes since the tutorial was written.

kdelwat / osm-ansible