Dubster

Runs unittests for each dub packages against latest dmd compiler.

How does it work?

There are two components, a server and at least one worker. The server checks github and code.dlang.org for releases and new packages. Based on that it creates jobs.

Each worker polls the server for a job. Once a job is returned it starts a docker skoppe/dubster-digger container to build DMD (if not already cached). It then uses the fresh dmd and starts a docker skoppe/dubster-dub container and starts building the package. After the build is complete the raw stdout/stderr is analysed for known error/success patterns, then all is pushed back to the server which will pipe it straight into mongo.

The next steps are

• getting feedback
• add/remove stuff to this list and prioritize
• add reporting in the server
• writing an api
• a web-app to poke around the dataset
• fix some of the failing builds due to missing library dependency
• adding dmd nightly (which was the whole purpose)
• notifications
• track memory consumption during build (so we can push heavy builds only to heavy workers)
• windows/os-x workers
• probably a lot more

Running a Dubster Worker

First of all, do a docker pull skoppe/dubster-dub and docker pull skoppe/dubster-digger .

Create a data container. Dubster will fire off additional containers and needs a way to share data between them.

docker create -v /gen --name dubsterdata skoppe/dubster /bin/true

(Note to self: /gen/dub-cache must explicitly be created. Find nice solution.)

Then start the container:

docker run -d --volumes-from dubsterdata --name dubsterworker -e "DOCKER_HOST=http://172.17.0.1:2375" skoppe/dubster --worker --serverHost=https://ghozadab.skoppe.nl --memory=2147483648

Note:

The worker needs a docker host that listens on a tcp port for http requests. /var/run/docker.sock is currently not supported. Try to start the docker daemon with -H tcp://172.17.0.1:2375.

(TODO: This is really hackish. Improve)

Deploy your own Dubster Server

Note: don't just copy and paste these commands in your shell, adjust them according to your needs.

mongo

docker create -v /mongodata --name mongodata mongo /bin/bash -c "echo Started mongo data container"

docker run --volumes-from mongodata -p 27017 --name mongoserver -d mongo

nginx-proxy

Place a file default with client_max_body_size 20m; as content somewhere (e.g. /root/nginx.conf/default)

Note: adjust /root/letscerts to your liking

docker run -d -p 80:80 -p 443:443 --name nginx-proxy -v /root/letscerts:/etc/nginx/certs:ro -v /usr/share/nginx/html -v /var/run/docker.sock:/tmp/docker.sock:ro -v /root/nginx.conf:/etc/nginx/vhost.d jwilder/nginx-proxy

letsencrypt-nginx-proxy-companion

docker run -d -v /root/letscerts:/etc/nginx/certs:rw --volumes-from nginx-proxy -v /var/run/docker.sock:/var/run/docker.sock:ro jrcs/letsencrypt-nginx-proxy-companion

Upon starting it will first setup keys etc. It might take 5 min before it will correctly route and all that.

dubster

Note: Use your own host and email :)

docker run -d --name dubsterserver -e "VIRTUAL_HOST=some.public.host" -e "LETSENCRYPT_HOST=some.public.host" -e "LETSENCRYPT_EMAIL=e@ma.il" --link mongoserver:mongo -e "MONGO_DB_NAME=dubby" skoppe/dubster --server

Development

CI

Wercker CI will build and push a new image to docker hub on each git push. (TODO: isolate docker hub push only to master branch)

If you want to fork and have your own, you need to setup wercker (or other) and adjust the container names appropriately.