[bug] End-to-end integration tests are failing

Describe the bug

All of the cloud provider end-to-end tests are currently failing with following error message:

[terraform]: │ Error: error initializing keycloak provider
[terraform]: │ 
[terraform]: │   with provider["registry.terraform.io/mrparkers/keycloak"],
[terraform]: │   on kubernetes.tf line 140, in provider "keycloak":
[terraform]: │  140: provider "keycloak" {
[terraform]: │ 
[terraform]: │ error sending POST request to
[terraform]: │ https://qhub-integration-test-aws.qhub.dev/auth/realms/master/protocol/openid-connect/token:
[terraform]: │ 401 Unauthorized
[terraform]: ╵

[danlester]

I’ve been using Digital Ocean and Google Cloud this week without any problems.

But I have always used HTTPS with genuine Let’s Encrypt certs.

While most of qhub deploy can work to a ‘fake’ domain, or one that isn’t really accessible locally (maybe due to outdated DNS or cert problems), the Terraform Keycloak provider needs to communicate directly with the Keycloak API on a valid domain.

This is the kind of problem that will result in ‘error initializing keycloak provider’ (the first one eskild reported above). The code aims to tell it not to verify TLS if the qhub-config.yaml specifies a self-signed cert or if it is a ‘local’ deployment - but it’s possible that’s the bit that’s broken as I’ve only done genuine Let’s Encrypt deployments recently.

As for your more recent message (Kubernetes cluster unreachable) that doesn’t look like a Keycloak problem as such - just a Helm one. But Keycloak is the first thing to be installed of course, so it’s likely that JupyterHub (and maybe everything else) would also fail for the same reason.

[iameskild]

Closed here https://github.com/Quansight/qhub-integration-test/pull/26