openstreetmap-tile-server

This container allows you to easily set up an OpenStreetMap PNG tile server given a .osm.pbf file. It is based on the latest Ubuntu 18.04 LTS guide from switch2osm.org and therefore uses the default OpenStreetMap style.

Setting up the server

First create a Docker volume to hold the PostgreSQL database that will contain the OpenStreetMap data:

docker volume create osm-data

Next, download an .osm.pbf extract from geofabrik.de for the region that you're interested in. You can then start importing it into PostgreSQL by running a container and mounting the file as /data/region.osm.pbf. For example:

docker run \
    -v /absolute/path/to/luxembourg.osm.pbf:/data/region.osm.pbf \
    -v osm-data:/data/database/ \
    overv/openstreetmap-tile-server \
    import

If the container exits without errors, then your data has been successfully imported and you are now ready to run the tile server.

Note that the import process requires an internet connection. The run process does not require an internet connection. If you want to run the openstreetmap-tile server on a computer that is isolated, you must first import on an internet connected computer, export the osm-data volume as a tarfile, and then restore the data volume on the target computer system.

Also when running on an isolated system, the default index.html from the container will not work, as it requires access to the web for the leaflet packages.

Automatic updates (optional)

If your import is an extract of the planet and has polygonal bounds associated with it, like those from geofabrik.de, then it is possible to set your server up for automatic updates. Make sure to reference both the OSM file and the polygon file during the import process to facilitate this, and also include the UPDATES=enabled variable:

docker run \
    -e UPDATES=enabled \
    -v /absolute/path/to/luxembourg.osm.pbf:/data/region.osm.pbf \
    -v /absolute/path/to/luxembourg.poly:/data/region.poly \
    -v osm-data:/data/database/ \
    overv/openstreetmap-tile-server \
    import

Refer to the section Automatic updating and tile expiry to actually enable the updates while running the tile server.

Please note: If you're not importing the whole planet, then the .poly file is necessary to limit automatic updates to the relevant region. Therefore, when you only have a .osm.pbf file but not a .poly file, you should not enable automatic updates.

Letting the container download the file

It is also possible to let the container download files for you rather than mounting them in advance by using the DOWNLOAD_PBF and DOWNLOAD_POLY parameters:

docker run \
    -e DOWNLOAD_PBF=https://download.geofabrik.de/europe/luxembourg-latest.osm.pbf \
    -e DOWNLOAD_POLY=https://download.geofabrik.de/europe/luxembourg.poly \
    -v osm-data:/data/database/ \
    overv/openstreetmap-tile-server \
    import

Using an alternate style

By default the container will use openstreetmap-carto if it is not specified. However, you can modify the style at run-time. Be aware you need the style mounted at run AND import as the Lua script needs to be run:

docker run \
    -e DOWNLOAD_PBF=https://download.geofabrik.de/europe/luxembourg-latest.osm.pbf \
    -e DOWNLOAD_POLY=https://download.geofabrik.de/europe/luxembourg.poly \
    -e NAME_LUA=sample.lua
    -e NAME_STYLE=test.style
    -e NAME_MML=project.mml
    -e NAME_SQL=test.sql
    -v /home/user/openstreetmap-carto-modified:/data/style/ \
    -v osm-data:/data/database/ \
    overv/openstreetmap-tile-server \
    import

If you do not define the "NAME_*" variables, the script will default to those found in the openstreetmap-carto style.

Be sure to mount the volume during run with the same -v /home/user/openstreetmap-carto-modified:/data/style/

If you do not see the expected style upon run double check your paths as the style may not have been found at the directory specified. By default, openstreetmap-carto will be used if a style cannot be found

Only openstreetmap-carto and styles like it, eg, ones with one lua script, one style, one mml, one SQL can be used

Running the server

Run the server like this:

docker run \
    -p 8080:80 \
    -v osm-data:/data/database/ \
    -d overv/openstreetmap-tile-server \
    run

Your tiles will now be available at http://localhost:8080/tile/{z}/{x}/{y}.png. The demo map in leaflet-demo.html will then be available on http://localhost:8080. Note that it will initially take quite a bit of time to render the larger tiles for the first time.

Using Docker Compose

The docker-compose.yml file included with this repository shows how the aforementioned command can be used with Docker Compose to run your server.

Preserving rendered tiles

Tiles that have already been rendered will be stored in /data/tiles/. To make sure that this data survives container restarts, you should create another volume for it:

docker volume create osm-tiles
docker run \
    -p 8080:80 \
    -v osm-data:/data/database/ \
    -v osm-tiles:/data/tiles/ \
    -d overv/openstreetmap-tile-server \
    run

If you do this, then make sure to also run the import with the osm-tiles volume to make sure that caching works properly across updates!

Enabling automatic updating (optional)

Given that you've set up your import as described in the Automatic updates section during server setup, you can enable the updating process by setting the UPDATES variable while running your server as well:

docker run \
    -p 8080:80 \
    -e REPLICATION_URL=https://planet.openstreetmap.org/replication/minute/ \
    -e MAX_INTERVAL_SECONDS=60 \
    -e UPDATES=enabled \
    -v osm-data:/data/database/ \
    -v osm-tiles:/data/tiles/ \
    -d overv/openstreetmap-tile-server \
    run

This will enable a background process that automatically downloads changes from the OpenStreetMap server, filters them for the relevant region polygon you specified, updates the database and finally marks the affected tiles for rerendering.

Tile expiration (optional)

Specify custom tile expiration settings to control which zoom level tiles are marked as expired when an update is performed. Tiles can be marked as expired in the cache (TOUCHFROM), but will still be served until a new tile has been rendered, or deleted from the cache (DELETEFROM), so nothing will be served until a new tile has been rendered.

The example tile expiration values below are the default values.

docker run \
    -p 8080:80 \
    -e REPLICATION_URL=https://planet.openstreetmap.org/replication/minute/ \
    -e MAX_INTERVAL_SECONDS=60 \
    -e UPDATES=enabled \
    -e EXPIRY_MINZOOM=13 \
    -e EXPIRY_TOUCHFROM=13 \
    -e EXPIRY_DELETEFROM=19 \
    -e EXPIRY_MAXZOOM=20 \
    -v osm-data:/data/database/ \
    -v osm-tiles:/data/tiles/ \
    -d overv/openstreetmap-tile-server \
    run

Cross-origin resource sharing

To enable the Access-Control-Allow-Origin header to be able to retrieve tiles from other domains, simply set the ALLOW_CORS variable to enabled:

docker run \
    -p 8080:80 \
    -v osm-data:/data/database/ \
    -e ALLOW_CORS=enabled \
    -d overv/openstreetmap-tile-server \
    run

Connecting to Postgres

To connect to the PostgreSQL database inside the container, make sure to expose port 5432:

docker run \
    -p 8080:80 \
    -p 5432:5432 \
    -v osm-data:/data/database/ \
    -d overv/openstreetmap-tile-server \
    run

Use the user renderer and the database gis to connect.

psql -h localhost -U renderer gis

The default password is renderer, but it can be changed using the PGPASSWORD environment variable:

docker run \
    -p 8080:80 \
    -p 5432:5432 \
    -e PGPASSWORD=secret \
    -v osm-data:/data/database/ \
    -d overv/openstreetmap-tile-server \
    run

Performance tuning and tweaking

Details for update procedure and invoked scripts can be found here link.

THREADS

The import and tile serving processes use 4 threads by default, but this number can be changed by setting the THREADS environment variable. For example:

docker run \
    -p 8080:80 \
    -e THREADS=24 \
    -v osm-data:/data/database/ \
    -d overv/openstreetmap-tile-server \
    run

CACHE

The import and tile serving processes use 800 MB RAM cache by default, but this number can be changed by option -C. For example:

docker run \
    -p 8080:80 \
    -e "OSM2PGSQL_EXTRA_ARGS=-C 4096" \
    -v osm-data:/data/database/ \
    -d overv/openstreetmap-tile-server \
    run

AUTOVACUUM

The database use the autovacuum feature by default. This behavior can be changed with AUTOVACUUM environment variable. For example:

docker run \
    -p 8080:80 \
    -e AUTOVACUUM=off \
    -v osm-data:/data/database/ \
    -d overv/openstreetmap-tile-server \
    run

FLAT_NODES

If you are planning to import the entire planet or you are running into memory errors then you may want to enable the --flat-nodes option for osm2pgsql. You can then use it during the import process as follows:

docker run \
    -v /absolute/path/to/luxembourg.osm.pbf:/data/region.osm.pbf \
    -v osm-data:/data/database/ \
    -e "FLAT_NODES=enabled" \
    overv/openstreetmap-tile-server \
    import

Warning: enabling FLAT_NOTES together with UPDATES only works for entire planet imports (without a .poly file). Otherwise this will break the automatic update script. This is because trimming the differential updates to the specific regions currently isn't supported when using flat nodes.

Benchmarks

You can find an example of the import performance to expect with this image on the OpenStreetMap wiki.

Troubleshooting

ERROR: could not resize shared memory segment / No space left on device

If you encounter such entries in the log, it will mean that the default shared memory limit (64 MB) is too low for the container and it should be raised:

renderd[121]: ERROR: failed to render TILE default 2 0-3 0-3
renderd[121]: reason: Postgis Plugin: ERROR: could not resize shared memory segment "/PostgreSQL.790133961" to 12615680 bytes: ### No space left on device

To raise it use --shm-size parameter. For example:

docker run \
    -p 8080:80 \
    -v osm-data:/data/database/ \
    --shm-size="192m" \
    -d overv/openstreetmap-tile-server \
    run

For too high values you may notice excessive CPU load and memory usage. It might be that you will have to experimentally find the best values for yourself.

The import process unexpectedly exits

You may be running into problems with memory usage during the import. Have a look at the "Flat nodes" section in this README.

License

Copyright 2019 Alexander Overvoorde

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

    https://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.