Standard setup for an e2e test environment on supported clusters is provided in this directory. Subdirectories contain the default profiles. The deployment uses the openshift/origin-ansible:latest
image to launch clusters.
Prerequisites:
- Docker 1.12+ installed and available on your local machine
- Your system time must be up to date in order for gcloud to authenticate to GCE (run
sudo ntpd -gq
in your VM if necessary)
gcp-dev
(default) - 1 master, 3 node cluster with bootstrapping (uses theopenshift-gce-devel
project, suitable for developers)gcp
- 1 master, 3 node cluster with bootstrapping (uses theopenshift-gce-devel-ci
project, suitable for CI)gcp-ha
- 3 master, 3 node cluster with bootstrapping (uses theopenshift-gce-devel-ci
project, suitable for CI)
In each profile, see vars-origin.yaml
for the default settings (which is common across all profiles) and vars.yaml
for the variables that differ.
This is the typical workflow to configure the development profile gcp-dev
(but would apply to any other profile).
Clone the shared secrets repo to $SHARED_SECRETS
:
git clone git@github.com:openshift/shared-secrets.git $SHARED_SECRETS
Copy secrets to their well-known locations in the profile directory:
cp $SHARED_SECRETS/gce/aos-serviceaccount.json $RELEASE_TOOLS/gcp-dev/gce.json
cp $SHARED_SECRETS/gce/cloud-user@gce.pem $RELEASE_TOOLS/gcp-dev/ssh-privatekey
cp $SHARED_SECRETS/mirror/ops-mirror.pem $RELEASE_TOOLS/gcp-dev/ops-mirror.pem
Add new installer variables to a file such as gcp-dev/zz_vars.yaml
(any YAML file in the profile directory will be treated as a variables file, and for now, the zz_
prefix ensures that your variables will override any predefined ones). Keep in mind that the precedence order of the variable files in the profile is unspecified: to override variables in vars.yaml
or vars-origin.yaml
just modify those files directly for now.
You can set an alternate profile by passing PROFILE=gcp
on the make
arguments.
# launch cluster using the latest RPMs
$ make WHAT=mycluster up
# after cluster is launched, admin.kubeconfig is copied locally
$ KUBECONFIG=gcp-dev/admin.kubeconfig oc status
# teardown cluster
$ make WHAT=mycluster down
# get a shell in the ansible environment
$ make WHAT=mycluster sh
# launch a different profile
$ make WHAT=mycluster PROFILE=gcp-ha
The following ansible variables are commonly used:
openshift_test_repo
is the URL to a yum repo to use to installopenshift_image_tag
(optional) is the suffix of the image to run withopenshift_pkg_version
(optional) is the exact package value to use-0.0.1
if installing RPMs that don't match the desired version
The following environment variables can be provided:
OPENSHIFT_ANSIBLE_IMAGE
(defaults toopenshift/origin-ansible:latest
) the image to deploy from
The following files can be added to a profile:
gce.json
containing the GCP service account credentials to use that have permission to create instances, networks, and a bucket in the project used by the profileops-mirror.pem
containing the client certificate for the OpenShift ops mirrorssh-privatekey
(optional) with the private key to use for the GCP instancesssh-privatekey.pub
(optional) a key to use with the above private key
-
Add the cluster's registry (e.g.
docker-registry-default.apps.$WHAT.origin-gce.dev.openshift.com
) to the local Docker daemon's insecure registry list. -
Log in to the cluster's registry using an SA from the namespace which should receive the image:
oc sa get-token -n openshift builder | docker login -u builder --password-stdin docker-registry-default.apps.$WHAT.origin-gce.dev.openshift.com
-
Tag an image pointing at the cluster registry in the desired namespace:
docker tag github.com/openshift/origin:b42d282 docker-registry-default.apps.$WHAT.origin-gce.dev.openshift.com/openshift/origin:b42d282
-
Push an image to the cluster registry:
docker push docker-registry-default.apps.$WHAT.origin-gce.dev.openshift.com/openshift/origin:b42d282