bookwyrm-kubernetes

This is a collection of files that I used to run Bookwyrm on Kubernetes.

Bookwyrm originally runs with Docker-compose, which is not super-scalable. The installation instruction is made with the assumption that some infrasctructure, such as redis and database, isn't already available.

My installation

The configuration files on this repo assume that I am:

• using only one redis instance for both activity and broker,
• using S3 for storage
• having an external reverse-proxy in addition to my nginx-ingress on my k8s cluster.
• using an already configured reverse proxy
• relying on an already present postgres install
• using a https proxy for the S3 files, so that all files are shown from my domain, and not directly from my S3 provider. I prefer this so that it is easier to migrate to another bucket later. A good example on how to do this is on Mastodon documentation: https://docs.joinmastodon.org/admin/optional/object-storage-proxy/

Installation procedure

• Clone Bookwyrm-repo:
  git clone git@github.com:bookwyrm-social/bookwyrm.git
• Checkout to the production branch: git checkout production
• build the image (You need to build the image and upload it to your container registry): docker buildx build --platform=linux/amd64 -t mycontainerregistry/bookwyrm:0.6.6 . --no-cache docker push mycontainerregistry/bookwyrm:0.6.6
• Configure your persistant volume (I used nfs) on k8s with the following folders:
• redis
• static
• images
• app
• Configure your nginx file, according to https://docs.joinbookwyrm.com/reverse-proxy.html. Usually, you copy /app/nginx/production to default.conf, and edit accordingly.
• Copy redis.conf to the root of your persistant volume
• Configure your configMap.yml and secrets.yml, as well as all the deployments, so that they point to your volume and container registry,
• After everything is configured, deploy these:

kubectl apply -f configMap.yml
kubectl apply -f secrets.yml
kubectl apply -f bookwyrm-service*
kubectl apply -f bookwyrm-ingress.yml
kubectl apply -f bookwyrm-worker.yml

• Get a shell on your worker pod, and execute the following commands:

python manage.py migrate
python manage.py migrate django_celery_beat
python manage.py initdb

• Now get the other deployments up:

kubectl apply -f bookwyrm-beat.yml
kubectl apply -f bookwyrm-flower.yml
kubectl apply -f bookwyrm-web.yml

• Finally, run the following commands on the web pod:

python manage.py compile_themes
python manage.py collectstatic --no-input
python manage.py admin_code

Considerations

I am not sure how easy it will be to update this on kubernetes, and I'm surely duplicating a lot of unnecessary work here. I believe things will be easier if I modify the image to actually copy the installation instead of mounting a volume for it, but I want to confirm this with the developers to make sure these files are purely static.

I had a bit of trouble with my instance as it was not respecting the USE_HTTPS: "true" configuration, so I created an .env file on my /app folder and added USE_HTTPS=true. It worked well after that.

It is most likely easier to maintain a Helm chart for this, but I'm not so familiar with Helm charts, so I decided to go for this simple setup.

The cool thing is that if my instance needs to grow, all I need is to increase the number of replicas for the woker nodes and for the web nodes.