eyra-k8s

Kubernetes configuration for EYRA deployment.

Deploying backend/frontend

Get the latest master build number from Travis (for frontend https://travis-ci.org/EYRA-Benchmark/eyra-frontend/builds).

Then, for frontend update tag line in eyra-chart/values.staging.yaml:

eyra-frontend:
  image:
    repository: eyra/frontend
    tag: 456 # <- update with the latest Travis build number

To update the backend:

web:
  debug: 'True'
  siteID: '2'
  imageName: 'eyra/comic:6' # <- update the number

After pushing to master, it should automatically update after 1 or 2 minutes.

Setting up k8s initially

NOTE: The cluster has helm 2.12.0 installed: make sure your helm version matches this!

• Create a Kubernetes cluster
• Allow ports 80 and 443 in the firewall for the cluster
• Add the kubectl config for the cluster to your ~/.kube/config
• Run preparation/setup.sh
• helm install --name eyra ./eyra-chart
• kubectl exec -it web-xxxxxx /bin/bash
• python manage.py migrate
• python manage.py check_permissions
• python manage.py create_demo_benchmark

Get a local shell connected to the cluster

• telepresence --run-shell
• If you want your env vars to be copied from a pods env, first copy the env, then export your local env back because it overwrites some values:

# run telepresence with sh, so that import/export is consistent
telepresence --run sh
OLD_ENV=$(export -p)
set -o allexport
$(kubectl exec eyra-dev-tom-web-5958f799b6-ng2cv sh -- -c env)
set +o allexport
eval $OLD_ENV

Use a local PyCharm dev server with the cluster

• ensure telepresence is installed
• telepresence --swap-deployment --expose 8000 --run /bin/bash
• Start PyCharm from the new bash terminal
• Open the backend (comic) project
• Run the dev server in PyCharm (no environment variables set)

Telepresence crash

One thing that can happen is that telepresence crashes (I haven't seen this), or that your machine falls off the network for any reason (bad wifi, laptop lid closed, etc.). Unfortunately, telepresence doesn't clean up in this case, leaving the original service unreachable. In this case you can do the following to restore the original deployment:

• kubectl delete svc,deploy -l telepresence
• kubectl scale --replicas=1 deploy [original deployment name]

Notes

When adding charts to requirements.yaml, retrieve the charts by calling::

helm dependency update

To update the helm configuration in the cluster after changing the helm chart (assuming you are in the root of this repo)::

helm upgrade -f ./eyra-chart/values.dev-<your name>.yaml eyra-dev-<your name> ./eyra-chart

OR: Apply the auto-updater script, this polls this github repository for updates and applies them every minute:

kubectl -n tiller create secret generic secrets-pass --from-literal=pass='secrets.zip_password'
kubectl apply -f preparation/auto-updater.yaml`

Docker registry listing::

curl -X GET https://docker-registry.roel.dev.eyrabenchmark.net/v2/_catalog
curl -X GET https://docker-registry.roel.dev.eyrabenchmark.net/v2/rzi/blabla/tags/list

Push image to Docker registry::

docker tag <image hash> docker-registry.roel.dev.eyrabenchmark.net/<repository>
docker push docker-registry.roel.dev.eyrabenchmark.net/<repository>

Update secrets.yaml (use password from LastPass)::

zip -e secrets.zip secrets.yaml

Running locally

Install minikube: https://kubernetes.io/docs/setup/minikube/#installation

Start minikube: minikube start

Install helm:

cd preparation
export TILLER_NAMESPACE=tiller
kubectl create namespace tiller
kubectl apply -f rbac-tiller.yaml
helm init --service-account tiller --tiller-namespace tiller
minikube addons enable ingress

cd ..
helm install ./eyra-chart --name eyra-dev --namespace eyra-dev -f ./eyra-chart/values.dev.yaml
# wait till done (might take a while, downloads images, prepares DB etc.

# get minikube IP:
minikube ip
# remember IP, e.g. 192.168.99.100
# add the following to /etc/hosts:
192.168.99.100 eyra.local
192.168.99.100 www.eyra.local
192.168.99.100 api.eyra.local
192.168.99.100 comic.eyra.local
192.168.99.100 docker-registry.eyra.local