This directory contains a number of Dockerfile for different configurations.
Each directory contains a ./build.sh for convenient building of the image.
Note: the mention of the overall licensing terms of the GDAL build is to the best of our knowledge and not guaranteed. Users should check by themselves.
Alpine version:
- 3.17 for GDAL 3.7
- 3.16 for GDAL 3.6
- 3.15 for GDAL 3.5
- Image size: ~ 59 MB
- Raster drivers: ultrasmall + built-in + SQLite-based ones + network-based ones
- Vector drivers: ultrasmall + built-in + most XML-based ones + network-based ones + PostgreSQL
- Using internal libtiff and libgeotiff
- External libraries enabled: ultrasmall + libexpat, libpq, libssl
- No GDAL Python
- Base PROJ grid package (http://download.osgeo.org/proj/proj-datumgrid-1.8.zip)
- Overall licensing terms of the GDAL build: permissive (MIT, BSD style, Apache, etc..)
- Image size: ~ 282 MB
- Raster drivers: small + netCDF, HDF5, BAG
- Vector drivers: small + Spatialite, XLS
- Using internal libtiff and libgeotiff
- External libraries enabled: small + libgeos, libhdf5, libhdf5, libkea, libnetcdf, libfreexl, libspatialite, libxml2, libkml, libpoppler, openexr, libheif, libdeflate, libparquet, libjxl
- GDAL Python
- Base PROJ grid package (http://download.osgeo.org/proj/proj-datumgrid-1.8.zip)
- Overall licensing terms of the GDAL build: copy-left (GPL) + LGPL + permissive
Ubuntu version:
- 22.04 for GDAL 3.6 and 3.7
- 20.04 for GDAL 3.4 and 3.5
- Image size: ~ 385 MB
- Raster drivers: all built-in + JPEG + PNG + JP2OpenJPEG + WEBP +SQLite-based ones + network-based ones
- Vector drivers: all built-in + XML based ones + SQLite-based ones + network-based ones + PostgreSQL
- Using internal libtiff and libgeotiff
- External libraries enabled: libsqlite3, libproj, libcurl, libjpeg, libpng, libwebp, libzstd, libexpat, libxerces-c, libpq, libssl, libgeos, libspatialite
- GDAL Python (Python 3.8 for Ubuntu 20.04, Python 3.10 for Ubuntu 22.04)
- Base PROJ grid package (http://download.osgeo.org/proj/proj-datumgrid-1.8.zip)
- Overall licensing terms of the GDAL build: LGPL + permissive (MIT, BSD style, Apache, etc..)
- Image size: ~ 1.48 GB
- Raster drivers: all based on almost all possible free and open-source dependencies
- Vector drivers: all based on almost all possible free and open-source dependencies
- Using internal libtiff and libgeotiff
- External libraries enabled: small + libnetcdf, libhdf4, libhdf5, libtiledb, libkea, mongocxx 3.4, libspatialite, unixodbc, libxml2, libcfitsio, libmysqlclient, libkml, libpoppler, pdfium, openexr, libheif, libdeflate, libparquet, libjxl
- GDAL Python (Python 3.8 for Ubuntu 20.04, Python 3.10 for Ubuntu 22.04)
- All PROJ grid packages (equivalent of latest of proj-data-X.zip from http://download.osgeo.org/proj/ at time of generation, > 500 MB)
- Overall licensing terms of the GDAL build: copy-left (GPL) + LGPL + permissive
Pull the required image and then run passing the gdal program you want to execute as a docker run command. Bind a volume from your local file system to the docker container to run gdal programs that accept a file argument. For example, binding -v /home:/home on Linux or -v /Users:/Users on Mac will allow you to reference files in your home directory by passing their full path. Use the docker --rm option to automatically remove the container when the run completes.
Note: you should not try to install GDAL (directly or indirectly through other packages that depend on it) with the package managing system (apt/apk) of the Linux distributions. It will conflict with the custom GDAL version provided by the Docker image and will likely result in a broken container.
docker pull ghcr.io/osgeo/gdal:alpine-small-latest
docker run --rm -v /home:/home ghcr.io/osgeo/gdal:alpine-small-latest gdalinfo $PWD/my.tifIf you are getting a <jemalloc>: arena 0 background thread creation failed (1) error message when running the osgeo/gdal[:ubuntu-full-XXXX] images on a Linux host with an old distribution (RHEL/CentOS 7), adding --privileged to the docker run command line should help (see OSGeo#6331)
Tagged images of recent past releases are available. The last ones (at time of writing) are for GDAL 3.7.1 and PROJ 9.2.1, for linux/amd64 and linux/arm64:
- ghcr.io/osgeo/gdal:alpine-small-3.7.1
- ghcr.io/osgeo/gdal:alpine-normal-3.7.1
- ghcr.io/osgeo/gdal:ubuntu-small-3.7.1
- ghcr.io/osgeo/gdal:ubuntu-full-3.7.1
Each directory contains a build.sh shell script that supports building images
for multiple platforms using an experimental feature called Docker BuildKit.
BuildKit CLI looks like docker buildx build vs. docker build
and allows images to build not only for the architecture and operating system
that the user invoking the build happens to run, but for others as well.
There is a small setup process depending on your operating system. Refer to Preparation toward running Docker on ARM Mac: Building multi-arch images with Docker BuildX.
If you're running Docker for MacOS with an Intel CPU
and you wanted to build the alpine-small image with support for Raspberry Pi 4,
adding a couple flags when running alpine-small/build.sh can greatly simplify this process
Use the two script flags in order to leverage BuildKit:
| Flag | Description | Arguments |
|---|---|---|
| --with-multi-arch | Will build using the buildx plugin |
N/A |
| --platform | Which architectures to build | linux/amd64,linux/arm64 |
Example
alpine-small/build.sh --with-multi-arch --release --gdal v3.2.0 --proj master --platform linux/arm64,linux/amd64
Override the base image, used to build and run gdal, by setting the environment variable: BASE_IMAGE
Example
BASE_IMAGE="debian:stable" ubuntu-small/build.sh --release --gdal v3.2.0 --proj master
Override the image and repository of the final image by setting the environment variable: TARGET_IMAGE
Example
TARGET_IMAGE="YOU_DOCKER_USERNAME/gdal" alpine-small/build.sh --release --gdal v3.2.0 --proj master