Skip to content

Latest commit

 

History

History
63 lines (50 loc) · 3.34 KB

docker.md

File metadata and controls

63 lines (50 loc) · 3.34 KB
Raw

Docker-based development

SplinterDB currently only supports Linux. If you develop on Mac or Windows (or just want to isolate your SplinterDB work from your host machine), you may use Docker to build SplinterDB from source and/or integrate an application against the SplinterDB library.

This document assumes some existing proficiency with Docker workflows.

Beware: While our CI system produces and uses the container images described here, the SplinterDB maintainers do not use these exact workflows for daily development, so there may be rough edges. Please report any issues (or submit pull-requests) and we'll do our best to keep this working.

Building SplinterDB from source in a Docker container

Our continuous-integration system publishes (and uses) a container image called build-env that has all build-time dependencies. It is built from Dockerfile.build-env and can be pulled directly via

docker pull projects.registry.vmware.com/splinterdb/build-env

To use it to build SplinterDB from source, change into the directory containing this repository and then do

$ docker run -it --rm --mount type=bind,source="$PWD",target=/splinterdb-src \
     projects.registry.vmware.com/splinterdb/build-env /bin/bash

Note: the build-env image contains a working build environment, but does not contain the SplinterDB source code. That must be mounted into the running container, e.g. the --mount command shown above.

Inside the container is a Linux environment with all necessary dependencies for building from source, with either GCC or Clang. From within there, you ought to be able to follow the steps in our build docs.

For example, from inside the running container:

docker$ cd /splinterdb-src
docker$ export CC=clang  # or gcc
docker$ export LD=clang
docker$ make
docker$ make run-tests
docker$ make install

Using pre-built SplinterDB

In addition to the build-environment described above, our continuous integration system also publishes a minimal container image called splinterdb that contains only the build outputs and dependencies for run-time usage of SplinterDB. It is built from this Dockerfile and may be pulled as

docker pull projects.registry.vmware.com/splinterdb/splinterdb

If you don't need to modify SplinterDB itself, this image should be sufficient for linking the SplinterDB library into another program, without bringing along all the build-time dependencies.

Example usage:

$ docker run -it --rm projects.registry.vmware.com/splinterdb/splinterdb /bin/bash

The container image includes:

  • Runtime dependencies for SplinterDB, including libaio and libxxhash
  • SplinterDB static and shared libraries: /usr/local/lib/libsplinterdb.{a,so}
  • Header files for SplinterDB's public API: /usr/local/include/splinterdb/
  • Some pre-built test binaries and test scripts: /splinterdb/bin/... and /splinterdb/test.sh

Note: the splinterdb image does not include tools to build SplinterDB itself from source. See above for that.

Consider using bind mounts to access your application source code from your host OS:

$ docker run -it --rm \
    --mount type=bind,source="$PWD/my-app-src",target=/my-app-src \
    projects.registry.vmware.com/splinterdb/splinterdb /bin/bash