New blog infrastructure
I finally got around to automating the publishing of this blog. It hasn’t been a high priority, since I only post a couple of times a year, but it’s always bothered me that something that is such an ideal candidate for automation hasn’t been automated. Anyway, I finally did it and I want to describe the setup in case it’s helpful for anyone looking to do the same.
The blog is published to Bitbucket pages (I used Bitbucket because it supports Mercurial and life is too short to use unpleasant tools in your hobby time), and I wrote my own static-site generator (this is not an advertisement; do not use it; it’s not remotely “production ready”) because at least at the time Hugo seemed overly complex (and thus tedious to learn) for my use-case and writing my own seemed like more fun.
I decided to put my sources in one repository and the generated output in the repository which would be published. The target repo is called weberc2.bitbucket.io–this naming convention is (or at least it was when I started my blog) the mechanism by which Bitbucket knew to serve your blog as a static site. The source repo is a private repo.
The workflow was to write a post in the source repo’s
posts/ directory, pull
down the latest changes to the target repo, clear out the target repo, run the
generator (copying output files into the empty target repo), and finally
commit/push the changes to both repos.
Instead, I wanted to be able to work against the source repo and have it automatically publish when I pushed changes. To do that, I updated my static site generator with a Bitbucket Pipeline (a CI pipeline tool integrated into Bitbucket) to build and publish Docker images.
pipelines: default: - step: services: - docker caches: - docker script: - docker build -t weberc2/neon:latest . - docker login --username $DOCKER_USERNAME --password $DOCKER_PASSWORD - docker tag weberc2/neon weberc2/neon:$BITBUCKET_COMMIT - docker push weberc2/neon:latest - docker push weberc2/neon:$BITBUCKET_COMMIT
I also created a pipeline in the source repository which uses that Docker image to run the static site generator and update the target repository.
pipelines: default: - step: services: - docker caches: - docker script: # NOTE that we're deliberately not passing `-it`; passing it yields # the error: "the input device is not a TTY" - docker run --rm -v $PWD:/site --workdir /site weberc2/neon neon build - hg clone ssh://firstname.lastname@example.org/weberc2/weberc2.bitbucket.org /weberc2.bitbucket.org - rm -rf /weberc2.bitbucket.org/* - cp -r ./_output/* /weberc2.bitbucket.org/ - cd /weberc2.bitbucket.org - hg addremove - hg commit -m "Automatic update from source.weberc2.bitbucket.org @ $BITBUCKET_COMMIT" - hg push
This took some work to allow the source repo’s pipeline to write to the target repo–basically I created an SSH key for the source pipeline and added its public key to my Bitbucket profile so the target repository would allow the pipeline to update it as though I were making the updates. This isn’t ideal, since I’m giving the pipeline access to my whole repo, but the alternative is to create a Bitbucket user for each pipeline (which I may eventually do). Ideally the target repo would allow me to assign permissions for the pubkey without having to map the pubkey to a Bitbucket user, but alas…
It only took me an hour or so to build this out. If you’re interested in doing something similar and/or have questions about my setup. Feel free to reach out to me on Twitter.
I want to take a moment to plug Bitbucket Pipelines. The user experience is nothing short of fantastic. The only issue I ran into was the permissioning bit, but that’s hardly insurmountable. The docs are excellent, the secrets management works as expected, the UI is intuitive, etc. I’ve been steeped in frustrating CI tools for the last 6 months, and Pipelines really stands out as a solid tool. Definitely the best thing I’ve found for getting a small project up and running quickly. I’m not getting any kickbacks for this paragraph; just giving credit where it’s due.