k8s-release-manager

README

Note: Release Manager is a community driven project. LogicMonitor support will not assist in any issues related to Release Manager.

Release Manager is a tool for importing and exporting Helm release state

• Export Helm release state: Release Manager can contact Tiller in the configured cluster, collect all metadata for each deployed release, and write that metadata to the configured backend. This metadata can later be consumed by Release Manager import to re-install the saved releases to a different cluster.

• Continuously export Helm release state: Release Manager can also be run in daemon mode to continuously update the stored state to reflect ongoing changes to the cluster. When running in daemon mode, it is HIGHLY recommended when running to use the official Release Manager Helm chart.

• Import Helm release state: Release Manager can retrieve stored state from the configured backend and install all exported releases to the current Kubernetes cluster.

• Configurable storage backends: Release Manager can interact with multiple release state storage backends. Currently supported backends are:

  • S3 (Note: S3 object versioning is strongly recommended for backend buckets)
  • Local (Note: the local backend is intended for non-production use only)

Release Manager Overview

Release Manager provides functionality for exporting and importing the state of Helm releases currently deployed to a Kubernetes cluster. The state of the installed releases is saved to a configurable backend for easy restoration of previously-deployed releases or for simplified re-deployment of those releases to a new Kubernetes cluster.

Release Manager operations can be run locally or within the Kubernetes cluster. The application also supports a daemon mode that will periodically update the saved state.

To export releases, Release Manager queries Tiller to collect metadata for all releases currently deployed in the source cluster and writes this metadata to the configured backend data store. If the Release Manager is deployed in daemon mode via its own Helm chart, it will also store metadata about itself. This metadata is used to prevent import operations from creating a new Release Manager with the same configuration as the previous managed, causing both instances to write conflicting state to the backend.

To import releases, Release Manager retrieves the state stored in the backend, connects to Tiller in the target Kubernetes cluster, and deploys the saved releases to the cluster.

Release Manager will use --kubeconfig/--kubecontext, $KUBECONFIG, or ~/.kube/config to establish a connection to the Kubernetes cluster. If none of these configuraitons are set, an in-cluster connection will be attempted. All actions will be performed against the current cluster and a given command will only perform actions against a single cluster, i.e. 'export' will export releases from the configured cluster while 'import' will deploy releases to the configured cluster and 'clear' requires no custer connection whatsoever.

Command usage

Installing via Helm Chart

Installing releasemanager daemon via Helm chart

helm repo add logicmonitor https://logicmonitor.github.io/k8s-helm-charts
helm install logicmonitor/releasemanager \
  --set path=$BACKEND_STORAGE_PATH \
  --name releasemanager-$CURRENT_CLUSTER

For detailed information about the Helm chart, see the README

Viewing installed releases

When running in daemon mode, the Release Manager exposes an endpoint to view the list of releases currently stored in the backend. This endpoint is available at /releases.

Note that if the Release Manager is running in-cluster, you'll need to expose its service via ingress using --set ingress.hosts={...} when installing the Helm chart.

Retrieving metrics

When running in daemon mode, the Release Manager exposes an endpoint providing monitoring metrics. This endpoint is available at /debug/vars at /debug/vars.

Note that if the Release Manager is running in-cluster, you'll need to expose its service via ingress using --set ingress.hosts={...} when installing the Helm chart.

Use case examples

Release Manager was created with the goal of solving two common cluster management problems. These use cases are outlined below along with a general instructions for using Release Manager to solve the problem.

Redeploying applications after a blue/green switch

Using blue/green deployments for Kubernetes clusters is a great upgrade strategy; however, when a cluster is running dozens or hundreds of application deployments supporting several different environments, it can become burdensome to redploy all of these applications every time a cluster is upgraded. To solve this problem, Release Manager makes it easy to take a snapshot of applications deployed to the current cluster and redeploy those applications to the new cluster.

1. Install Release Manager locally

go get github.com/logicmonitor/k8s-release-manager/cmd/releasemanager

2. Export the releases currently installed in the source cluster

releasemanager export local \
  --kubeconfig $SOURCE_CLUSTER_KUBECONFIG \
  --path $LOCAL_RELEASE_STATE_PATH

3. Deploy the saved releases to the destination cluster

releasemanager import local \
  --kubeconfig $DESTINATION_CLUSTER_KUBECONFIG \
  --path $LOCAL_RELEASE_STATE_PATH

Deploying applications in a disaster recovery scenario

Some disaster scenarios can result in the need to create a brand new Kubernetes cluster for failover. When a cluster is running dozens or hundreds of application deployments, it can become burdensome and incur huge time costs to redploy all of these applications manually. To solve this problem, Release Manager can operate as a long-running process inside your cluster to periodically export snapshots of deployed applications to an external storage backend. In the event of a disaster recovery operation requiring a cluster failover, Release Manager can retrieve the stored state and quickly deploy all of the applications from the failed cluster to the new cluster.

1. Add the LogicMonitor Helm repository

helm repo add logicmonitor https://logicmonitor.github.io/k8s-helm-charts

2. Install Release Manager locally

go get github.com/logicmonitor/k8s-release-manager/cmd/releasemanager

3. Deploy Release Manager in daemon mode when provisioning a production cluster

helm install logicmonitor/releasemanager \
  --set path=$PROD_CLUSTER_BACKEND_PATH \
  --set s3.bucket=$RELEASE_MANAGER_STATE_BUCKET \
  --set s3.region=$BUCKET_REGION \
  --name releasemanager-$PROD_CLUSTER_NAME

4. Provision a failover cluster during a disaster scenario

5. Deploy saved releases to the failover cluster.

NOTE! Be sure to --new-path. This prevents Release Manager in failover cluster from overwriting the production state

releasemanager import s3 \
  --kubeconfig $FAILOVER_CLUSTER_KUBECONFIG \
  --path $PROD_CLUSTER_BACKEND_PATH \
  --new-path $FAILOVER_CLUSTER_BACKEND_PATH \
  --bucket $RELEASE_MANAGER_STATE_BUCKET \
  --region $BUCKET_REGION

License