digitalocean-cloud-controller-manager

Kubernetes Cloud Controller Manager for DigitalOcean

digitalocean-cloud-controller-manager is the Kubernetes cloud controller manager implementation for DigitalOcean. Read more about cloud controller managers here. Running digitalocean-cloud-controller-manager allows you to leverage many of the cloud provider features offered by DigitalOcean on your Kubernetes clusters.

Releases

Cloud Controller Manager follows semantic versioning. Although the version is still below v1, the project is considered production-ready.

Because of the fast Kubernetes release cycles, CCM (Cloud Controller Manager) will only support the version that is also supported on DigitalOcean Kubernetes product. Any other releases will be not officially supported by us.

Getting Started

Learn more about running DigitalOcean cloud controller manager here!

Note that this CCM is installed by default on DOKS (DigitalOcean Managed Kubernetes), you don’t have to do it yourself.

Examples

Here are some examples of how you could leverage digitalocean-cloud-controller-manager:

• loadbalancers
• node labels and addresses

Production notes

do not modify DO load-balancers manually

When you are creating load-balancers through CCM (via LoadBalancer-typed Services),it is very important that you must not change the DO load-balancer configuration manually. Such changes will eventually be reverted by the reconciliation loop built into CCM. There is one exception in load-balancer name which can be changed (see also the documentation on load-balancer ID annotations).

Other than that, the only safe place to make load-balancer configuration changes is through the Service object.

DO load-balancer entry port restrictions

For technical reasons, the ports 50053, 50054, and 50055 cannot be used as load-balancer entry ports (i.e., the port that the load-balancer listens on for requests). Trying to use one of the affected ports as a service port causes a 422 entry port is invalid HTTP error response to be returned by the DO API (and surfaced as a Kubernetes event).

The solution is to change the service port to a different, non-conflicting one.

Development

Basics

• Go: min v1.25.x (see go.mod file)

This project uses Go modules for dependency management and employs vendoring. Please ensure to run make vendor after any dependency modifications.

After making your code changes, run the tests and CI checks:

make ci

Run Locally

If you want to run digitalocean-cloud-controller-manager locally against a particular cluster, keep your kubeconfig ready and start the binary in the main package-hosted directory like this:

cd cloud-controller-manager/cmd/digitalocean-cloud-controller-manager
REGION=fra1 DO_ACCESS_TOKEN=your_access_token go run main.go \
  --kubeconfig <path to your kubeconfig file>                     \
  --leader-elect=false --v=5 --cloud-provider=digitalocean

The REGION environment variable takes a valid DigitalOcean region. It can be set to keep digitalocean-cloud-controller-manager from trying to access the DigitalOcean metadata service which is only available on droplets. If the REGION variable is set, then the DO Regions service will be used to validate the specified region. It can also be set for local development purposes. Overall, which region you choose should not matter a lot as long as you pick one.

You might also need to provide your DigitalOcean access token in DO_ACCESS_TOKEN environment variable. The token does not need to be valid for the cloud controller to start, but in that case, you will not be able to validate integration with DigitalOcean API.

Please note that if you use a Kubernetes cluster created on DigitalOcean, there will be a cloud controller manager running in the cluster already, so your local one will compete for API access with it.

Optional features

Add Public Access Firewall

You can have digitalocean-cloud-controller-manager manage a DigitalOcean Firewall that will dynamically adjust rules for accessing NodePorts: once a Service of type NodePort is created, the firewall controller will update the firewall to public allow access to just that NodePort. Likewise, access is automatically retracted if the Service gets deleted or changed to a different type.

Example invocation:

cd cloud-controller-manager/cmd/digitalocean-cloud-controller-manager
DO_ACCESS_TOKEN=<your_access_token>                           \
PUBLIC_ACCESS_FIREWALL_NAME=firewall_name                     \
PUBLIC_ACCESS_FIREWALL_TAGS=worker-droplet                    \
digitalocean-cloud-controller-manager                         \
  --kubeconfig <path to your kubeconfig file>                 \
  --leader-elect=false --v=5 --cloud-provider=digitalocean

The PUBLIC_ACCESS_FIREWALL_NAME environment variable defines the name of the firewall. The firewall is created if no firewall by that name is found.

The PUBLIC_ACCESS_FIREWALL_TAGS environment variable refers to the tags associated with the droplets that the firewall should apply to. Usually, this is a tag attached to the worker node droplets. Multiple tags are applied in a logical OR fashion.

In some cases, firewall management for a particular Service may not be desirable. One example is that a NodePort is supposed to be accessible over the VPC only. In such cases, the Service annotation kubernetes.digitalocean.com/firewall-managed can be used to selectively exclude a given Service from firewall management. If set to "false", no inbound rules will be created for the Service, effectively disabling public access to the NodePort. (Note the quotes that must be included with “boolean” annotation values.) The default behavior applies if the annotation is omitted, is set to "true", or contains an invalid value.

No firewall is managed if the environment variables are missing or left empty. Once the firewall is created, no public access other than to the NodePorts is allowed. Users should create additional firewalls to further extend access.

Expose Prometheus Metrics

If you are interested in exposing Prometheus metrics, you can pass in a metrics endpoint that will expose them. The command will look similar to this:

cd cloud-controller-manager/cmd/digitalocean-cloud-controller-manager
DO_ACCESS_TOKEN=your_access_token                  \
METRICS_ADDR=<host>:<port>                         \
digitalocean-cloud-controller-manager              \
  --kubeconfig <path to your kubeconfig file>      \
  --leader-elect=false --v=5 --cloud-provider=digitalocean

The METRICS_ADDR environment variable takes a valid endpoint that you’d like to use to serve your Prometheus metrics. To be valid it should be in the form <host>:<port>.

After you have started up digitalocean-cloud-controller-manager, run the following curl command to view the Prometheus metrics output:

curl <host>:<port>/metrics

Admission Server

The admission server is an optional component aiming at reducing bad config changes for DO managed objects (LBs, etc). If you want to know more about it, read the docs.

DO API rate limiting

DO API usage is subject to certain rate limits. In order to protect against running out of quota for extremely heavy regular usage or pathological cases (e.g., bugs or API thrashing due to an interfering third-party controller), a custom rate limit can be configured via the DO_API_RATE_LIMIT_QPS environment variable. It accepts a float value, e.g., DO_API_RATE_LIMIT_QPS=3.5 to restrict API usage to 3.5 queries per second.

Run Containerized

If you want to test your changes in a containerized environment, create a new image with the version set to dev:

VERSION=dev make publish

This will create a binary with version dev and docker image pushed to digitalocean/digitalocean-cloud-controller-manager:dev.

Release a new version

Update Go and dependencies

1. Update Go dependencies

   go get -u ./...
   go mod tidy
   go mod vendor
   

2. Update Go version to latest GA version

Github Action (preferred)

To create the docker image and generate the manifests, go to the actions page on GitHub and click “Run Workflow”. Specify the GitHub <tag> that you want to create, making sure it is prefixed with a v. Running the workflow also requires that you temporarily turn off “Require a pull request before merging” setting in the master branch protection rules settings. Don’t forget to turn it back on once the release is done!

The workflow does the following:

• Runs make bump-version with <tag>
• Creates the ccm related manifests file as <tag>.yaml
• Commits the manifest file under releases/ directory in the repo
• Creates release and tags the new commit with the <tag> specified when workflow is triggered
• Logs in with dockerhub credentials specified as secrets
• Builds the docker image digitalocean/digitalocean-cloud-controller-manager:<tag>
• Pushes digitalocean/digitalocean-cloud-controller-manager:<tag> to dockerhub

Manual (deprecated)

NOTE: this workflow is deprecated, please prefer the Github Action workflow described above.

To manually release a new version, first bump the version:

make NEW_VERSION=v1.0.0 bump-version

Make sure everything looks good. Create a new branch with all changes:

git checkout -b release-<new version> origin/master
git commit -a -v
git push origin release-<new version>

After it’s merged to master, tag the commit and push it:

git checkout master
git pull
git tag <new version>
git push origin <new version>

Finally, create a Github release from master with the new version and publish it:

make publish

This will compile a binary containing the new version bundled in a docker image pushed to digitalocean/digitalocean-cloud-controller-manager:<new version>

Contributing

At DigitalOcean we value and love our community! If you have any issues or would like to contribute, feel free to open an issue/PR and cc any of the maintainers below.