ichiong / drone-k8s

drone CI plugin for executing rolling updates to kubernetes

drone-k8s plugin

Plugin for Drone CI to be used in the publish/deploy steps that will perform a kubernetes rolling update of the appropriate replication controller and associated pod(s).

To use the plugin you will need to clone the repo, build the image, push it to your associated registry, and then use that as the image for the plugin as shown below. The easiest way is to use the provided Makefile. Run $ make to build and tag the images.

To use this plugin yourself you would just need to replace image: your-repo/your-org/drone-k8s:1.0.0 with the appropriate values.

If updating a kubernetes deployment you can specify that with is_deployment. The default behavior is to perform a rolling-update of a replication controller if is_deployment is not set or false.

Commands we use to update a deployment

# perform patch of deployment which enlists the "strategy" defined in the resource definition
kubectl patch ... my-deployment -p `{ ... }`

This plugin assumes your drone server is running inside of kubernetes.

Available Plugin Options

replication_controller   -- REQUIRED for rolling-update: The name of the rc
docker_image             -- REQUIRED for both: The image name with appropriate docker repo
service_config_map_path  -- REQUIRED if updating a config map 
namespace                -- the k8s namespace (defaults to using `default`)
k8s_service_host         -- the K8S_SERVICE_HOST env var (default is 10.100.0.1)
k8s_service_port         -- the K8S_SERVICE_PORT env var (default is 443)
protocol                 -- https || http   (default is https)
path_to_cert_authority   -- absolute path to the cert authority (ca.pem)
path_to_client_key       -- absolute path to the client key (worker-key.pem)
path_to_client_cert      -- absolute path to the client cert (worker.pem)
update_period            -- (only used for rolling-update) the update period for the rolling update (default is 1m0s)
timeout                  -- (only used for rolling-update) the timeout threshold for the rolling update (default is 5m0s)
is_deployment            -- REQUIRED for deployment update: Is this an update of a deployment or not. If not specified then rolling-update of replication controller is assumed.
container_name           -- REQUIRED for deployment update or if performing rolling-update of multi-container pod: The name of the container to update the image with.
deployment_resource_name -- REQUIRED for deployment update: The name of the deployment resource (i.e. my-deployment)
esb_config_path          -- REQUIRED for any Geofeedia service with a ConfigMap configured ESB : very specific to Geofeedia... sorry no more pure OSS :(
config_map_name          -- REQUIRED for any Geofeedia service with a ConfigMap configured ESB : very specific to Geofeedia... sorry no more pure OSS :(
config_map_key_name      -- REQUIRED for any Geofeedia service with a ConfigMap configured ESB : very specific to Geofeedia... sorry no more pure OSS :(
service_config_map_path  -- REQUIRED for any Geofeedia service with a ConfigMap (this will be the file that is used in the `kubectl replace -f <CONFIGMAP_FILE_HERE>`)

Examples

# perform a rolling-update
publish: 
  drone-k8s:
    image: your-repo/your-org/drone-k8s:1.0.0
    replication_controller: some-rc
    namespace: some-ns
    docker_image: some-repo/some-org/some-image:1.0.0
    path_to_cert_authority: /path/to/ca.pem
    path_to_client_key: /path/to/worker-key.pem
    path_to_client_cert: /path/to/worker.pem
    update_period: 5s
    timeout: 30s
    
# perform a strategic update for a deployment along with a replace of a ConfigMap
publish: 
  drone-k8s:
    image: your-repo/your-org/drone-k8s:1.0.0
    namespace: some-ns
    is_deployment: true
    deployment_resource_name: some-deployment
    container_name: some-container
    service_config_map_path: /some/path/to/my-configmap.yaml
    docker_image: some-repo/some-org/some-image:1.0.0
    path_to_cert_authority: /path/to/ca.pem
    path_to_client_key: /path/to/worker-key.pem
    path_to_client_cert: /path/to/worker.pem

Also, if you want to test this locally (replacing the "vargs" section with whatever you want to pass in as params for the plugin) you can do...

$ go run main.go <<EOF
{
    "repo": {
        "clone_url": "git://github.com/drone/drone",
        "owner": "drone",
        "name": "drone",
        "full_name": "drone/drone"
    },
    "system": {
        "link_url": "https://beta.drone.io"
    },
    "build": {
        "number": 22,
        "status": "success",
        "started_at": 1421029603,
        "finished_at": 1421029813,
        "message": "Update the Readme",
        "author": "johnsmith",
        "author_email": "[email protected]",
        "event": "push",
        "branch": "master",
        "commit": "436b7a6e2abaddfd35740527353e78a227ddcb2c",
        "ref": "refs/heads/master"
    },
    "workspace": {
        "root": "/drone/src",
        "path": "/drone/src/github.com/drone/drone"
    },
    "vargs": {
        "is_deployment": "true",
        "deployment_resource_name": "my-deployment",
        "container_name": "my-container",
        "docker_image": "quay.io/geofeedia/image:tag",
        "namespace": "my-namespace"
    }
}
EOF