Docker is a virtualized environment running a Docker demon, in which you can run software without altering your host system permanently. The software components run in containers that are easy to setup and tear down individually. The Docker demon can use operating-system-level virtualization (Linux, Windows) or a virtual machine (macOS, Windows).
This allows to set up a development environment for OpenStreetMap Carto easily. Specifically, this environment consists of a PostgreSQL database to store the OpenStreetMap data and Kosmtik for previewing the style.
Docker is available for Linux, macOS and Windows. Install the software packaged for your host system in order to be able to run Docker containers. You also need Docker Compose, which should be available once you installed Docker itself. Otherwise you need to install Docker Compose manually.
You need sufficient disk space of several Gigabytes. Docker creates a disk image for its virtual machine that holds the virtualised operating system and the containers. The format (Docker.raw, Docker.qcow2, *.vhdx, etc.) depends on the host system. It can be a sparse file allocating large amounts of disk space, but still the physical size starts with 2-3 GB for the virtual OS and grows to 6-7 GB when filled with the containers needed for the database, Kosmtik, and a small OSM region. Further 1-2 GB are needed for shape files in the openstreetmap-carto/data repository.
If you are eager to get started here is an overview over the necessary steps. Read on below to get the details.
git clone https://github.com/gravitystorm/openstreetmap-carto.gitto clone openstreetmap-carto repository into a directory on your host system- download OpenStreetMap data in osm.pbf format to a file
data.osm.pbfand place it within the openstreetmap-carto directory (for example some small area from Geofabrik) - If necessary,
sudo service postgresql stopto make sure you don't have currently running a native PostgreSQL server which would conflict with Docker's PostgreSQL server. docker-compose up importto import the data (only necessary the first time or when you change the data file). Additionally you can set import options through environment variables. More on that laterdocker-compose up kosmtikto run the style preview application- browse to http://127.0.0.1:6789 to view the output of Kosmtik
- Ctrl+C to stop the style preview application
docker-compose stop dbto stop the database container
Instructions above will clone main OpenStreetMap Carto repository. To test your own changes you should fork gravitystorm/openstreetmap-carto repository and clone your fork.
This OpenStreetMap Carto repository needs to be a directory that is shared between your host system and the Docker virtual machine. Home directories are shared by default; if your repository is in another place you need to add this to the Docker sharing list (e.g. macOS: Docker Preferences > File Sharing; Windows: Docker Settings > Shared Drives).
OpenStreetMap Carto needs a database populated with rendering data to work. You first need a data file to import.
It's probably easiest to grab an PBF of OSM data from Geofabrik.
Once you have that file put it into the openstreetmap-carto directory and run docker-compose up import in the openstreetmap-carto directory.
This starts the PostgreSQL container (downloads it if it not exists) and starts a container that runs osm2pgsql to import the data. The container is built the first time you run that command if it not exists.
At startup of the container the script scripts/docker-startup.sh is invoked which prepares the database and itself starts osm2pgsql for importing the data. Then the scripts/get-external-data.py is called to download and import needed shapefiles.
osm2pgsql has a few command line options and the import by default uses a RAM cache of 512 MB, 1 worker and expects the import file to be named data.osm.pbf. If you want to customize any of these parameters you have to set the environment variables OSM2PGSQL_CACHE (e.g. export OSM2PGSQL_CACHE=1024 on Linux to set the cache to 1 GB) for the RAM cache (the value depends on the amount of RAM you have available, the more you can use here the faster the import may be), OSM2PGSQL_NUMPROC for the number of workers (this depends on the number of processors you have and whether your harddisk is fast enough e.g. is a SSD), or OSM2PGSQL_DATAFILE if your file has a different name.
You can also tune the PostgreSQL during the import phases, with PG_WORK_MEM (default to 16MB) and PG_MAINTENANCE_WORK_MEM (default to 256MB), which will eventually write work_mem and maintenance_work_mem to the postgresql.auto.conf once, making them applied each time the database started. Note that unlike osm2pgsql variables, once thay are set, you can only change them by running ALTER SYSTEM on your own, changing postgresql.auto.conf or remove the database volume by docker-compose down -v && docker-compose rm -v and import again.
get-external-data.py script has option -C (--cache) to save data after download (useful when you are tinkering with docker and ending up deleting volumes).
It also has option --no-update to stop program from downloading newer versions of shapefiles if you don't deem updating them necessary. Best used in conjunction with -C.
If everything goes out of the window, option --force will forcefully download data and import it. Option --force-import will try to force just import part.
Use EXTERNAL_DATA_SCRIPT_FLAGS env variable to pass those options. For example:
EXTERNAL_DATA_SCRIPT_FLAGS="--cache --no-update"
will keep data you downloaded and not update them (saving you on bandwidth) until you change this options.
If you want to customize and remember the values, supply it during your first import:
PG_WORK_MEM=128MB PG_MAINTENANCE_WORK_MEM=2GB \
OSM2PGSQL_CACHE=2048 OSM2PGSQL_NUMPROC=4 \
OSM2PGSQL_DATAFILE=taiwan.osm.pbf \
EXTERNAL_DATA_SCRIPT_FLAGS="--cache --no-update" \
docker-compose up import
Note that on Linux you need to export those environment variables before calling docker-compose. If you are using sudo to call docker (because your user is not in the docker group (which we don't recommend)), you need to also use sudo -E option
Variables will be remembered in .env if you don't have that file, and values in the file will be applied unless you manually assign them. Keep in mind this means if you change your .env file, but keep your environment variables untouched (you haven't unset them or you haven't rebooted your host), they will be used instead of anything that you changed in .env.
Depending on your machine and the size of the extract the import can take a while. When it is finished you should have the data necessary to render it with OpenStreetMap Carto.
After you have the necessary data available you can start Kosmtik to produce a test rendering. For that you run docker-compose up kosmtik in the openstreetmap-carto directory. This starts a container with Kosmtik and also starts the PostgreSQL database container if it is not already running. The Kosmtik container is built the first time you run that command if it not exists.
At startup of the container the script scripts/docker-startup.sh is invoked which downloads necessary shapefiles with scripts/get-external-data.py (if they are not already present). It afterwards runs Kosmtik. If you have to customize anything, you can do so in the script. The Kosmtik config file can be found in .kosmtik-config.yml.
If you want to have a local configuration for our project.mml you can place a localconfig.js or localconfig.json file into the openstreetmap-carto directory.
The shapefile data that is downloaded is owned by the user with UID 1000. If you have another default user id on your system, consider changing the line USER 1000 in the file Dockerfile.
After startup is complete you can browse to http://127.0.0.1:6789 to view the output of Kosmtik. By pressing Ctrl+C on the command line you can stop the container. The PostgreSQL database container is still running then (you can check with docker ps). If you want to stop the database container as well you can do so by running docker-compose stop db in the openstreetmap-carto directory.
Importing the data needs a substantial amount of RAM in the virtual machine. If you find the import process (Reading in file: data.osm.pbf, Processing) being killed by the Docker demon, exiting with error code 137, increase the Memory assigned to Docker (e.g. macOS: Docker Preferences / Windows: Docker Settings > Advanced > Adjust the computing resources).
Docker copies log files from the virtual machine into the host system, their location depends on the host OS. E.g. the 'console-ring' appears to be a ringbuffer of the console log, which can help to find reasons for killings.
While installing software in the containers and populating the database, the disk image of the virtual machine grows in size, by Docker allocating more clusters. When the disk on the host system is full (only a few MB remaining), Docker can appear stuck. Watch the system log files of your host system for failed allocations.
Docker stores its disk image by default in the home directories of the user. If you don't have enough space here, you can move it elsewhere. (E.g. macOS: Docker > Preferences > Disk).
When working with the style's database tables after an import, it can be helpful to log in at the console to inspect the table structure or view imported data. The following command will open a psql console on the database:
docker-compose exec -e PGUSER=postgres -e PGDATABASE=gis db psql