witw-chart

Helm chart for Posthog WITW

Overview

There are three charts in this repository:

• caddy, which is used to deploy a stateful (just for certs) reverse-proxy for our services
• postgis, which is used to deploy a stateful postgresql instance with the postgis extension
• witw, which is used to deploy stateless services based on the "whereintheworld" image

Deploying it

To deploy the full application, these steps should be followed (in order, DB first):

Postgres

1. First, cd into postgis and run helm template .
2. Does it look "correct enough" to you, all the environment variables, etc? If so, continue
3. Then, run helm install postgis .
4. This will render and upload the resource YAMLs to Kubernetes
5. After a minute, run kubectl get pods and make sure you see the postgis-0 pod in Running

Whereintheworld

1. First, cd into witw and run helm template . -f values/backend.yaml
2. (The -f values/backend.yaml tells it to use the "backend" values file and settings)
3. Does it look "correct enough" to you, all the environment variables, etc? If so, continue
4. Then, run helm install witw-backend .
5. This will render and upload the resource YAMLs to Kubernetes
6. After a minute, run kubectl get pods and make sure you see a witw-backend pod in Running
7. Then, follow the same steps but with -f values/frontend.yaml for the frontend!

Migrations

1. Run kubectl get pods and find the name of a witw-backend pod
2. Run kubectl exec -it $WITW_POD -- sh to get a shell into the running pod above
3. Run python manage.py migrate in that shell to kick off the migrations
4. If it tells you "No migrations to apply", then you're all up to date!

Caddy

1. First, cd into caddy and run helm template .
2. Does it look "correct enough" to you? If so, continue
3. Then, run helm install caddy .
4. This will render and upload the resource YAMLs to Kubernetes
5. After a minute, run kubectl get pods and make sure you see the caddy-0 pod in Running
6. You can also run kubectl logs caddy-0 to see how the TLS cert process went

Testing it out

After all of the above steps have been taken, try running:

curl -v https://hog.willett.io/_health

You should get a 200 response with server: gunicorn in the headers (and a 300 if you try with http). Additionally, if you try with a path looking like /api/foo, you should get a 400 error with "unauthorized".

Modifying it

For both of the charts mentioned above, the process for deploying changes is the same:

1. Make any changes to the chart you like
2. Run helm diff upgrade $RELEASE_NAME . to see what Helm would do if you actually applied it
3. Make a PR with your changes, include the diff if you like, get it approved
4. Run helm upgrade $RELEASE_NAME . to actually execute your update, make sure it works
5. Merge your PR right after

Poking around

Here are a few tools you might need when poking around the cluster:

• To get a (dangerous!) postgres shell: kubectl exec -it postgis-0 -- psql -U whereintheworld
• To run Django scripts: kubectl get pods then kubectl exec -it $WITW_POD -- python manage.py foo
• To hit the API from your laptop: kubectl port-forward svc/witw-backend 8000:8000 (then curl)