v0.7.2

[Release 0.7.2]

Note: This is a bug fix release.

When we bump the etcd version to 3.2, the images were only available on gcr.io . But now it is added on quay.io . We'd better use quay.io and keep it compatible to work for 3.1 versions of etcd.

v0.7.1

[Release 0.7.1]

NOTE: This release requires Kubernetes 1.8 .

Added

• TLS etcd cluster feature for EtcdBackup
• Log collector program for collecting logs in e2e test.
• ClusterSpec: In PodPolicy, add generic Affinity field to substitute bool AntiAffinity field.
• ClusterSpec: Add Repository field to substitute BaseImage field.

Changed

• Default etcd repository (base image) is changed to gcr.io/etcd-development/etcd, default etcd version is 3.2.11.
• Migrate dependency management tooling from glide to dep.
• Containerize e2e test in a pod instead of running on raw jenkin slave.

Removed

• ClusterSpec: Remove PodPolicy.AutomountServiceAccountToken field.
  No etcd pod will have service account token mounted.

Fixed

• Ignore Terminating pods when polling etcd pods.

Deprecated

• ClusterSpec: BaseImage is deprecated. It will be automatically converted to Repository in this release.
• ClusterSpec: In PodPolicy, AntiAffinity is deprecated. It will be automatically converted to Affinity.PodAntiAffinity
  terms with label selector on given cluster name and topology key on node in this release.

Security

• All operator images by default uses user etcd-operator instead of root.

v0.7.0

[Release 0.7.0]

Existing backup and restore features in EtcdCluster API won’t be supported after 0.7.0 release.
See Decoupling Backup and Restore Logic from Etcd Operator for more detail.

If applicable then see the upgrade guide on how to upgrade from v0.6.1 to v0.7.0 .

Added

• Add ServiceName and ClientPort into ClusterStatus.
• Add etcd backup operator for backing up an etcd cluster.
• Add etcd restore operator for restoring an etcd cluster.

Removed

• Remove pv-provisioner flag from etcd operator.
• Remove etcd cluster Backup feature from etcd operator.
• Remove etcd cluster Restore from etcd operator.

v0.6.1

[Release 0.6.1]

The operator will no longer create a storage class specified by --pv-provisioner by default. If applicable then see the upgrade guide on how to upgrade from v0.6.0 to v0.6.1 .

Added

• backup binary supports serving backup defined by backupSpec. In addition, when backupSpec
  is specified, backup binary changes to serve http backup requests only mode.
• Add operator flag --create-crd. By default it is true and operator will create EtcdCluster CRD.
  It can be set to false and operator won't create EtcdCluster CRD.
• Add operator flag --create-storage-class. By default it is false and operator won't create default storage class.
  It can be set to true and operator will create default storage class.

Changed

• An EtcdCluster CR with an invalid spec will not be marked as failed. Any changes that result in an invalid spec will be ignored and logged by the operator.

Fixed

• Fix the problem that operator might keep failing on version conflict updating CR status.

Deprecated

• The operator flag --pv-provisioner is depercated. We recommend to use per cluster storageClass.

v0.6.0

[Release 0.6.0]

BREAKING CHANGE: operator level S3 backup is removed. See upgrade on how to upgrade from 0.5.x to 0.6.0.

Added

• Add cluster events into EtcdCluster custom resource. See doc/user/conditions_and_events.md .

Changed

• Redefine status.conditions. See doc/user/conditions_and_events.md .

Removed

• Remove operator level S3 flag.
• Remove analytics flag. Disable Google analytics.

v0.5.2

[Release 0.5.2]

Added

• Expose /metrics endpoint at port 8080
• Add cluster S3 spec prefix feature. Let user choose a prefix under the bucket.
• Add automountServiceAccountToken to pod policy. Let users disable automounting of the Kubernetes access token into etcd-operator controlled pods.
• Cluster backups can now be saved using Azure Blob Storage (ABS).

Deprecated

• Deprecate operator S3 flag. Add warning note for using it in this release. The flag will be removed in 0.6.0 release.

v0.5.1

[Release 0.5.1]

Upgrade notice for TLS cluster users:
If you are using TLS-enabled etcd cluster, the SAN domain has been changed. See TLS docs. Before upgrading operator, you need to rotate certs on each secrets to allow both the old and new domains. Then restart each etcd pod -- the simplest way is to "upgrade" cluster version. Finally, it is safe to upgrade operator. It's highly recommended to save a backup before upgrade.

Added

• A new StorageClass spec field, allowing more granular control over how etcd clusters are backed up to PVs.

Changed

• Default timeout for snapshots done by backup sidecar increased from 5 seconds to 1 minute

Fixed

• Fix periodFullGC only executed once problem.
• [GH-1021] Use the cluster domain provided by kubelet instead of hardcoded .cluster.local .

v0.5.0

[Release 0.5.0]

BREAKING CHANGE: The cluster object will now be defined via a Custom Resource Definition(CRD) instead of a Third Party Resource(TPR). See the Changed section below for details.

Changed

• With k8s 1.7 and onwards TPRs have been deprecated and are replaced with CRD. See the k8s 1.7 blogpost or release notes for more details. For this release a live migration of the cluster spec from TPR to CRD is not supported. To preserve the cluster state during the upgrade you will need to create a backup of the cluster and recreate the cluster from the backup after upgrading the operator. See the upgrade guide for more detailed steps on how to do that.

• Changes in the cluster object's type metadata:

  • The apiVersion field has been changed from etcd.coreos.com/v1beta1 to etcd.database.coreos.com/v1beta2
  • The kind field has been changed from Cluster to EtcdCluster

v0.4.2

[Release 0.4.2]

Added

• GH-1232 the operator can now log critical actions like pod creation/deletion to a user specified path via the optional flag debug-logfile-path. The logs will only be generated if the cluster is self hosted and the flag is set. This can be used in conjunction with a persistent volume to persist the critical actions to disk for later inspection.

Changed

• enable alpha feature "tolerate unready endpoints" on etcd client and peer service. See alpha features doc for more details.

Fixed

• Fix append-hosts init-container not run on some restart cases.

v0.4.1

[Release 0.4.1]

This is mostly a bug-fix release.
We have done a lot of testing against k8s 1.7 and making it stable on 1.7 .

Added

• New self-hosted field SkipBootMemberRemoval allows users to skip the
  auto-deletion of the boot etcd member.

Fixed

• Make sure etcd pod's FQDN is resolvable before running etcd commands .