Occasionally, you may wish to spin up a separate k8s cluster for testing, say, infra stuff.
To do this:
# assuming you want to use the existing DB
./scripts/production/gke-create-cluster --with-new-database=false
If the hardcoded cluster version (of gke) is no longer available, that script will tell you what versions are currently supported.
Once this is done, you can build containers locally and deploy them to the cluster:
./scripts/deployment/_gcp-build-containers
./scripts/deployment/_gcp-push-images-to-gcr
./scripts/deployment/gke-deploy --cluster=darkcluster1570566003 --skip-rollbar
Note: the cluster argument can be had by running kubectl config get-contexts;
the value you want is darkcluster<timestamp> (the current prod cluster is
darkcluster). You can also get this cluster name from
https://console.cloud.google.com/kubernetes/list?project=balmy-ground-195100.
You may also want to edit
services/bwd-deployment/bwd-deployment.template.yaml and
services/editor-deployment/editor-deployment.template.yaml and
set replicas: 1, or something slightly less ridiculous than 72. (Deploys
will be faster, and you likely won't have prod-level traffic.)
If all has gone well, you now have a cluster ... on a different load balancer.
Get the IP address (kubectl get ingress), and then use curl with --resolve
to make certs DTRT: curl --resolve darklang.com:443:ip https://darklang.com/
(hostname, obviously, may be other things depending what you are testing).
When done, go to https://console.cloud.google.com/kubernetes/list?project=balmy-ground-195100 and delete your new cluster manually.