Addon Operator coordinates the lifecycle of Addons in managed OpenShift.
All development tooling can be accessed via make
, use make help
to get an overview of all supported targets.
This development tooling is currently used on Linux amd64, please get in touch if you need help developing from another Operating system or architecture.
To contribute new features or add/run tests, podman
or docker
and the go
tool chain need to be present on the system.
Dependencies are loaded as required and are kept local to the project in the .cache
directory and you can setup or update all dependencies via make dependencies
Updating dependency versions at the top of the Makefile
will automatically update the dependencies when they are required.
If both docker
and podman
are present you can explicitly force the container runtime by setting CONTAINER_RUNTIME
.
e.g.:
CONTAINER_RUNTIME=docker make dev-setup
Before making your first commit, please consider installing pre-commit and run pre-commit install
in the project directory.
Pre-commit is running some basic checks for every commit and makes our code reviews easier and prevents wasting CI resources.
Just wanting to play with the operator deployed on a cluster?
# In checkout directory:
make test-setup
This command will:
- Setup a cluster via kind
- Install OLM and OpenShift Console
- Compile your checkout
- Build containers
- Load them into the kind cluster (no registry needed)
- Install the Addon Operator
This will give you a quick environment for playing with the operator.
You can also use it to develop integration tests, against a complete setup of the Addon Operator:
# edit tests
# Run all integration tests and skip setup and teardown,
# as the operator is already installed by: make test-setup
make test-integration-short
# repeat!
The make
tooling offers the following flags to tweak your local in-cluster installation of the AddonOperator:
ENABLE_WEBHOOK=true/false
: Deploy the AddonOperator webhook server that runs Admission webhooksWEBHOOK_PORT=<PORT>
: Port to use while running the webhook serverENABLE_API_MOCK=true/false
: Deploy the mock OCM API server for testing and validating the UpgradePolicy flowENABLE_MONITORING=true/false
: Deploy the kube-prometheus monitoring stack for adding / testing AddonOperator metrics
To iterate fast on code changes and experiment, the operator can also run out-of-cluster. This way we don't have to rebuild images, load them into the cluster and redeploy the operator for every code change.
Prepare the environment:
make dev-setup
This command will:
- Setup a cluster via kind
- Install OLM and OpenShift Console
# Install Addon Operator CRDs
# into the cluster.
make setup-addon-operator-crds
# Make sure we run against the new kind cluster.
export KUBECONFIG=$PWD/.cache/integration/kubeconfig
# Set Addon operator namespace environment variable
export ADDON_OPERATOR_NAMESPACE=addon-operator
# Run the operator out-of-cluster:
# Mind your `KUBECONFIG` environment variable!
make run-addon-operator-manager
To run all the unit test and mock test into the local system. This way we can test the testable parts of the operator, individually and independently for proper operation.
# To run all the unit tests and mock tests
make test-unit
Warning:
- Your code runs as
cluster-admin
, you might run into permission errors when running in-cluster. - Code-Generators need to be re-run and CRDs re-applied via
make setup-addon-operator-crds
when code under./apis
is changed.
When using docker to spin a new Kind cluster, kube-proxy
would not start throwing this error:
I0511 11:47:28.965997 1 conntrack.go:100] Set sysctl 'net/netfilter/nf_conntrack_max' to XXXXXX
F0511 11:47:28.966114 1 server.go:495] open /proc/sys/net/netfilter/nf_conntrack_max: permission denied
Make sure to:
sudo sysctl net/netfilter/nf_conntrack_max=<value>
, and add a drop-in file to /etc/sysctl.d/99-custom.conf
to set the kernel parameters permanently.
The AddonOperator is instrumented with the prometheus-client provided by controller-runtime to record some useful Addon metrics.
Metric name | Type | Description |
---|---|---|
addon_operator_addons_count |
GaugeVec |
Total number of Addon installations, grouped by 'available', 'paused' and 'total' |
addon_operator_paused |
Gauge |
A boolean that tells if the AddonOperator is paused (1 - paused; 0 - unpaused) |
addon_operator_ocm_api_requests_durations |
Summary |
OCM API request latencies in microseconds. Grouped using tail-latencies (p50, p90, p99) |
addon_operator_addon_health_info |
GaugeVec |
Addon Health information (0 - Unhealthy; 1 - Healthy; 2 - Unknown) |
See Quickstart for instructions on how to setup a local monitoring stack for development / testing.
# 1. edit VERSION file
vim VERSION
# 2. update packaging files and prepare release
./mage prepare_release
# 3. create a new PR in openshift/addon-operator
# 4. create a Github Release/Git Tag on the latest commit of the main branch
Addon Operator is deployed via app-interface. Example MR updating staging: https://gitlab.cee.redhat.com/service/app-interface/-/merge_requests/36630
Example MR updating production: https://gitlab.cee.redhat.com/service/app-interface/-/merge_requests/36743