M3DB Operator

Project Status: Alpha

The M3DB Operator is a project dedicated to setting up M3DB on Kubernetes. It aims to automate everyday tasks around managing M3DB. Specifically, it aims to automate:

• Creating M3DB clusters
• Destroying M3DB clusters
• Expanding clusters (adding instances)
• Shrinking clusters (removing instances)
• Replacing failed instances

More information:

• Documentation
• Gitter room
• M3DB Google Group

Getting Started

The following instructions serve as a quickstart to get an M3DB cluster up and running in your Kubernetes cluster. This setup is not for production use, as there's no persistent storage. More information on production-grade clusters can be found in our docs.

Kubernetes Cluster Prerequisites

The M3DB operator targets Kubernetes 1.10 and 1.11. We generally aim to target the latest two minor versions supported by GKE but welcome community contributions to support more versions!

The M3DB operator is intended for creating highly available clusters across distinct failure domains. For this reason we currently only support Kubernetes clusters with nodes in at least 3 zones, but support for zonal clusters is coming soon.

When running on GKE, the user applying the manifests will need the ability to allow cluster-admin-binding during the installation. Use the following ClusterRoleBinding with the user name provided by gcloud:

kubectl create clusterrolebinding cluster-admin-binding --clusterrole=cluster-admin --user=<[email protected]>

Installing the M3DB Operator

With Helm:

helm repo add m3db https://s3.amazonaws.com/m3-helm-charts-repository/stable;
helm install m3db/m3db-operator --namespace m3db-operator

With kubectl (will install in the default namespace):

kubectl apply -f https://raw.githubusercontent.com/m3db/m3db-operator/v0.2.0/bundle.yaml

Managing Clusters

Creating a Cluster

Create a simple etcd cluster to store M3DB's topology:

kubectl apply -f https://raw.githubusercontent.com/m3db/m3db-operator/v0.2.0/example/etcd/etcd-basic.yaml

Apply manifest with your zones specified for isolation groups:

apiVersion: operator.m3db.io/v1alpha1
kind: M3DBCluster
metadata:
  name: simple-cluster
spec:
  image: quay.io/m3db/m3dbnode:latest
  replicationFactor: 3
  numberOfShards: 256
  # Default endpoints if using provided etcd manifests.
  etcdEndpoints:
  - http://etcd-0.etcd:2379
  - http://etcd-1.etcd:2379
  - http://etcd-2.etcd:2379
  isolationGroups:
  - name: group1
    numInstances: 1
    nodeAffinityTerms:
    - key: failure-domain.beta.kubernetes.io/zone
      values:
      - <zone-a>
  - name: group2
    numInstances: 1
    nodeAffinityTerms:
    - key: failure-domain.beta.kubernetes.io/zone
      values:
      - <zone-b>
  - name: group3
    numInstances: 1
    nodeAffinityTerms:
    - key: failure-domain.beta.kubernetes.io/zone
      values:
      - <zone-c>
  podIdentityConfig:
    sources: []
  namespaces:
    - name: metrics-10s:2d
      preset: 10s:2d
  dataDirVolumeClaimTemplate:
    metadata:
      name: m3db-data
    spec:
      accessModes:
      - ReadWriteOnce
      resources:
        requests:
          storage: 100Gi

Resizing a Cluster

To resize a cluster, specify the new number of instances you want in each zone either by reapplying your manifest or using kubectl edit. The operator will safely scale up or scale down your cluster.

Deleting a Cluster

Delete a cluster using kubectl delete. You will to remove the etcd data as well, or wipe the data generated by the operator if you intend to reuse the etcd cluster for another M3DB cluster:

kubectl exec etcd-0 -- env ETCDCTL_API=3 etcdctl del --keys-only --prefix ""

Contributing

We welcome community contributions to to the M3DB operator! Please see CONTRIBUTING.md for more information. Please note that on creating a pull request you will be asked to agree to the Uber CLA before we can accept your contribution.

License

This project is licensed under the Apache license -- see the LICENSE file for details.