[Release 0.9.4]

Added

• Added spec.BackupSource.S3.ForcePathStyle to EtcdBackup to force path style s3 uploads. #2036
• Added spec.RestoreSource.S3.ForcePathStyle to EtcdRestore to force path style s3 downloads. #2036

Changed

• Update Go version to 1.11.5
• Update k8s to 1.12.6
• EtcdBackup: Support periodic backups. This change added 3 new fields to EtcdBackup schema, 2 variables in spec, 1 variables in status.
  • in spec.backupPolicy
    • maxBackup: maximum number of backups to keep.
    • backupIntervalInSecond: how often to perform backup operation.
  • in status
    • LastSuccessDate: last time to succeed in taking backup

Fixed

• Fixed a bug where same CR names in different namespaces with cluster-wide operator were not working as expected #2026
• Fixed a bug where cluster names could exceed 63 octets the maximum defined by RFC 1035 section 2.3.4 resulting in hanging pods #2027.

[Release 0.9.3]

Added

• Added spec.pod.DNSTimeoutInSecond to EtcdCluster that allows setting a maximum allowed time for the init container of the etcd pod to reverse DNS lookup its IP given the hostname.

Changed

• Update Go version to 1.11.2
• Update k8s to 1.11.4
• k8s codegen updates are longer performed via container. Go dependencies are now vendored and updates are performed with shell script locally.

Fixed

• Fixed leaking http connections while verifying backup snapshots. #1976

[Release 0.9.2]

Added

• Added the field spec.pod.securityContext to EtcdCluster that allows setting a specific PodSecurityContext for the etcd pods. #1949

Changed

• Update Go version to 1.10
• Build gcr.io/coreos-k8s-scale-testing/etcd-operator-builder:0.4.1-2 container with Go 1.10 and dep 0.4.1
• etcd pod containers no longer run with a non-root security context by default. This setting can be configured per cluster via the PodPolicy.

[Release 0.9.1]

Added

• Added optional flag --cluster-wide to etcd-operator to allow it to manage etcd clusters across all namespaces. #1777
• Added support for annotation etcd.database.coreos.com/scope: clusterwide in EtcdCluster to allow it to be managed by a cluster wide operator. #1777
• Added the field spec.pod.busyboxImage to the PodPolicy of the EtcdCluster to allow overriding the default busybox image used for the etcd pod's init container. #1928

Fixed

• Fixed a bug where the informer watch stream would timeout after 30s of not receiving an event. #1936

[Release 0.9.0]

Same as v0.8.4. The version is bumping to 0.9.0 due to adding a new ABS backup API into etcd-backup-operator.

[Release 0.8.4]

Added

• Added ABS support for backup and restore
• Added tag to initContainer to enable offline deploys
• Enabled configurable backup timeout in backup operator

Changed

• Set 30s default request timeout for kube client
• Change check-dns init container image to busybox:1.28.0-glibc to fix nslookup failure in some environment.

Removed

• Removed self-hosted code

[Release 0.8.3]

Added

• Added the option to use PersistentVolume as non-stable storage for etcd pods. This feature is still alpha and subject to change in future releases #1861

[Release 0.8.3]

Added

• Added the option to use PersistentVolume as non-stable storage for etcd pods. This feature is still alpha and subject to change in future releases #1861

Changed

• Changed etcd pod member names to be unique by having a random suffix instead of a sequence number. This change is backward compatible and should not affect operator upgrade. Previously the etcd pod names would look like:

  NAME                            READY     STATUS    RESTARTS   AGE
  example-etcd-cluster-0000       1/1       Running   0          1m
  example-etcd-cluster-0001       1/1       Running   0          1m
  example-etcd-cluster-0002       1/1       Running   0          1m
  

  After this change:

  NAME                                  READY     STATUS    RESTARTS   AGE
  example-etcd-cluster-2885zjw9he       1/1       Running   0          1m
  example-etcd-cluster-gghrmbeid4       1/1       Running   0          1m
  example-etcd-cluster-w5q9sn37fd       1/1       Running   0          1m
  

Fixed

• Fixed a bug where the restore operator would fail to restore the seed member because recreating an etcd pod with the same name as a recently deleted one would conflict as the older pod and its resources, like the DNS name, might still not be deleted. #1825

[Release 0.8.2]

Added

• Add support for backup and restore from custom S3 endpoint.

Changed

• All etcd pod containers now run as non-root.

[Release 0.8.1]

Bug fix release for etcd-restore-operator. Also unify versioning for all three components.

Changed

• etcd-restore-operator will create a service for itself as the backup storage proxy. Delete the service in deployment yaml.

Fixed

• Fix etcd-restore-operator wouldn't report error and keep looping if EtcdRestore name is different than EtcdCluster name.

[Release 0.8.0]

Important Changes

Both etcd backup operator and etcd restore operator have changed their CR definition. Please follow the latest backup/restore CR definition for future backup and restore.

Added

• Add readiness probe to etcd pod. The readiness state will be reflected on status.members.ready/unready.
• TLS etcd cluster support in backup/restore-operator.
• Add spec validation in restore operator.
• Add BackupStorageType to EtcdRestore.RestoreSpec to indicate type of the backup storage which is used as RestoreSource and validation of BackupStorageType in restore operator.
• Add EtcdClusterRef to EtcdRestore.RestoreSpec to reference an EtcdCluster resource whose metadata and spec will be used to create the new restored EtcdCluster CR.
• Add create-crd flag to etcd backup operator allowing user to disable automatic backup CRD creation.
• Add create-crd flag to etcd restore operator allowing user to disable automatic restore CRD creation.
• Add EtcdVersion and EtcdRevision to EtcdBackup.BackupStatus.
• BackupStatus: Add detailed error when backup fails.

Changed

• Rename BackupCRStatus to BackupStatus.
• EtcdBackup: BackupSpec passes in S3BackupSource.Path as the S3 path to save the backup.
• EtcdBackup: BackupSpec spec uses etcd endpoints to retrieve snapshot.
• Change default etcd version to 3.2.13.

Removed

• EtcdBackup: BackupSpec removed ClusterName field in favor of etcd endpoints.
• EtcdCluster: ClusterSpec removed deprecated BaseImage field.