Serverless deploy error "no token in bearer response"

Summary

Deployment of our Knative Ruby App fails at the deploy stage using all default settings as outlined in this documentation.

Steps to reproduce

  1. Fork our Knative Ruby App project or clone the files into a new repository.
  2. Set up a Kubernetes cluster and integrate it with GitLab.
  3. Install Helm Tiller.
  4. Install Knative and provide it with a domain name.
  5. Direct a wildcard A record to the IP address provided by Knative.
  6. Run a pipeline.
  7. Observe the error message in the job log (example below).

Example Project

This was reproduced in this project.

What is the current bug behavior?

Currently, the app is not able to be deployed due to an error in the deploy stage. The build stage encounters no issues.

What is the expected correct behavior?

That the app is deployed to the cluster sucessfully and an entry appears in the Serverless section of the project.

Relevant logs and/or screenshots

Output of the job log for the deploy stage showing the error.

Running with gitlab-runner 11.8.0-rc1 (7a7b8375)
  on docker-auto-scale 0277ea0f
Using Docker executor with image gcr.io/triggermesh/tm@sha256:e3ee74db94d215bd297738d93577481f3e4db38013326c90d57f873df7ab41d5 ...
Pulling docker image gcr.io/triggermesh/tm@sha256:e3ee74db94d215bd297738d93577481f3e4db38013326c90d57f873df7ab41d5 ...
Using docker image sha256:6b3f6590a9b30bd7aafb9573f047d930c70066e43955b4beb18a1eee175f6de1 for gcr.io/triggermesh/tm@sha256:e3ee74db94d215bd297738d93577481f3e4db38013326c90d57f873df7ab41d5 ...
Running on runner-0277ea0f-project-10060955-concurrent-0 via runner-0277ea0f-srm-1550181942-eccf4d7f...
Cloning repository...
Cloning into '/builds/tristan/kube-kube-kube'...
Checking out 692997e7 as master...
Skipping Git submodules setup
$ echo "$CI_REGISTRY_IMAGE"
registry.gitlab.com/tristan/kube-kube-kube
$ tm -n "$KUBE_NAMESPACE" --config "$KUBECONFIG" deploy service "$CI_PROJECT_NAME" --from-image "$CI_REGISTRY_IMAGE" --wait
Deployment started. Run "tm -n kube-kube-kube-10060955 describe service kube-kube-kube" to see the details
Waiting for ready state.2019/02/14 22:07:51 Revision "kube-kube-kube-00001" failed with message: "no token in bearer response:\n{\"errors\":[{\"code\":\"DENIED\",\"message\":\"access forbidden\"}],\"http_status\":403}".
ERROR: Job failed: exit code 1

Output of checks

This bug happens on GitLab.com

Results of GitLab environment info

GitLab.com 11.8.0-rc4-ee

It was suggested in this SO question that the image defined in the deploy stage may be the issue. However, a change to that leads to a different error.

/cc @danielgruesso

ZD (Internal): https://gitlab.zendesk.com/agent/tickets/114285