Skip to content


Switch branches/tags

Name already in use

A tag already exists with the provided branch name. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Are you sure you want to create this branch?


Maintained by a former CoreOS engineer and inspired from the etcd-operator designed for Kubernetes, the etcd-cloud-operator automatically bootstraps, monitors, snapshots and recovers etcd clusters on cloud providers.

Used in place of the etcd binary and with minimal configuration, the operator handles the configuration and lifecycle of etcd, based on data gathered from the cloud provider and the status of the etcd cluster itself.

In other words, the operator operator is meant to help human operators sleep at night, while their mysterious etcd data store keeps running safely, even in the event of process, instance, network, or even availability-zone wide failures.


  • Resize: By abstracting cluster management, resizing the cluster becomes straightforward as the underlying auto-scaling group can simply be scaled as desired.

  • Snapshots: Periodically, snapshots of the entire key-value space are captured, from each of the etcd members and uploaded to an encrypted external storage, allowing the etcd (or human) operator to restore the store at a later time, in any etcd cluster or instance.

  • Failure recovery: Upon failure of a minority of the etcd members, the managed members automatically restarts and rejoins the cluster without breaking quorum or causing visible downtime - First by simply trying to rejoin with their existing data set, otherwise trying to join as a new member with a clean state, or by replacing the entire instance if necessary.

  • Disaster recovery: In the event of a quorum loss, consequence of the simultaneous failure of a majority of the members, the operator coordinates to snapshot any live members and cleanly stop then, before seeding a new cluster from the latest data revision available once the expected amount of instances are ready to start again.

  • ACL support: A user can configure the ACL of etcd by providing an init-acl config in the config file. See for more information.

  • JWT auth token support: JWT auth token can be enabled by specifying the jwt-auth-token-config in the config file, similar to the etcd -auth-token flag. The JWT auth token is HIGHLY recommended for production deployment, especially when the init-acl config is also enabled, the JWT auth token can help avoid the potential invalid auth token issue.

The operator and etcd cluster can be easily configured using a YAML file. The configuration notably includes clients/peers TLS encryption/authentication, with the ability to automatically generate self-signed certificates if encryption is desired but authentication is not.

How to try it?

Running a managed etcd cluster using the operator is simply a matter of running the operator binary in a supported auto-scaling group (as of today, AWS and Kubernetes only).

  • Docker: Head over to docs/docker-testing for a single-line local 3-nodes deployment.

  • AWS: A Terraform module is available to easily bring up production-grade etcd clusters managed by the operator within AWS.

  • Kubernetes: A basic Helm chart is available to quickly get started with a 3-nodes StatefulSet.