Release Stork v2.3.2

Improvements

• Added purgeDeletedResources option to the Migration spec, allowing resources that have been deleted on the source cluster to be purged from the destination cluster. This is disabled by default (#522)
• Portworx: Added support for in-place cloudsnap restores. Supported with Portworx Enterprise v2.3.2 onwards (#505)

Bug Fixes

• A failure during snapshot object creation can no longer lead to a stuck GroupVolumeSnapshot (#494)
• Updated base container and packages to fix vulnerabilities detected by CVEs (#503)
• Fixes for storkctl (#504):
  • Added the missing backupName param when creating ApplicationRestore objects
  • Added the replacePolicy param when creating ApplicationClone and ApplicationRestore objects
  • Fixed an alias clash for the VolumeSnapshotRestore and ApplicationRestore objects

Docker Hub Image: openstorage/stork:2.3.2

Release Stork v2.3.1

Improvements

• Optimized some operations for migration and migrationschedule to reduce Kubernetes API calls (#483)
• Added checks when collecting the following resources for migration and backup (#488)
  • Secrets: Ignore auto-created secrets with well-known prefix names
  • ServiceAccounts: Ignore auto-created service accounts unless they have non-default image pull secrets configured
  • Ingress: Only collect namespace-scoped objects
  • Role/RoleBinding: Ignore auto-created objects starting with "system:"

Bug Fixes

• Auto-created resources no longer risk getting stuck when being applied during migration (#488)

Docker Hub Image: openstorage/stork:2.3.1

Release Stork v2.3.0

New Features

• Added the VolumeSnapshotRestore CRD, which performs snapshot restores in-place
  • This CRD can restore both regular volume snapshots and group volume snapshots
  • You can find instructions here (#353)
• Added support for backing-up and restoring applications (#372)
  • Backup and restore operations trigger for the volume first, followed by the Kubernetes resources
  • A BackupLocation CRD specifies the location of the backup
  • An ApplicationRestore object must refer to an ApplicationBackup that it wants to restore from
  • You can configure an ApplicationBackupSchedule object to perform periodic backups (#379)
  • Backups sync from a BackupLocation if not already present on the cluster and if the sync option is set to true in the BackupLocation
  • You can find instructions here
• Added support to clone applications between namespaces (#373)
  • You can only create ApplicationClone objects in the admin namespace
  • ClusterRole is not copied since it is a cluster scoped object
  • ClusterRoleBinding is merged to have the same binding for both namespaces
  • You can find instructions here
• Added new storkctl commands for the new CRDs

Improvements

• Added events for pods in case of scheduling failure from the extender (#338)
• Added adminClusterPair field in the Migration spec. You can use this to specify a cluster pair from the admin namespace which can be used to migrate cluster scoped resources (#361)
• Added validation for name and namespace in the clusterpair generated with storkctl generate clusterpair (#422)
• The migration-admin-namespace parameter for stork has been deprecated. Please use admin-namespace instead. This can be used to specify the admin namespace for Migration as well as the namespace where ApplicationClones can be created (#485)
• The ImagePullSecrets field in the default ServiceAccount will now be migrated (#460)
• Portworx
  • Added support for all operations for PVCs created with CSI driver (#366)
  • Add idempotency to local snapshot calls to handle crashes while a snapshot is being taken (#380)

Bug Fixes

• Fixed an issue in the health monitor by ignoring duplicate node IPs which would cause pods to be deleted from nodes where storage was online (#448)
• The kube apiserver sometimes returns Unauthorized when running on the cloud if there are temporary auth errors. Migration now retries on Unauthorized errors (#478)
• Default RBAC objects are now skipped during migration. This fixes issues where migration would attempt to delete and re-create cluster-admin clusterrole on the destination (#478)
• Fixed an issue that occurred during migration when an application didn't have annotations. This issue caused failures when trying to add the replica count as an annotation (#467)

Docker Hub Image: openstorage/stork:2.3.0

Release Stork v2.2.5

Improvements:

• Users can now set the annotation stork.libopenstorage.org/preferLocalNodeOnly to true on their pods if they want pods to be scheduled only on nodes where data is located (#414)
• Added support to migrate the following additional resources:
  • Templates (#405)
  • Roles (#440)
  • RoleBindings (#440)
  • Ingress (#440)

Bug Fixes

• Ignore errors when updating DeploymentConfig using storkctl activate/deactivate migrations if the type isn't registered (#395)
• Fixed issue with failing scheduled snapshots if the snapshot schedule name was longer than 64 characters (#415)
• Fixed an issue where Portworx driver would attempt to make API calls to a non-existing portworx service port (#407)

Docker Hub Image: openstorage/stork:2.2.5

Release Stork v2.2.4

Enhancements

• Added a new status SyncStatusUnknown for ClusterDomains (#396)
• Migrations will be marked as Failed if the local cluster domain is inactive (#397)
• Updated migration behavior for PVs (#397)
  • Setting reclaim policy to Retain if volumes aren't being migrated
  • Updating PV if it already exists instead of deleting and creating

Bug Fixes

• Portworx: Changed the GetClusterDomains API to not return an error when it fails to fetch the volume list (#396)

Docker Hub Image: openstorage/stork:2.2.4

Release Stork v2.2.3

Improvements

• Optimized migrations for clusterrole and clusterrolebindings (#381)
  • For clusterrolebindings using the object that we are iterating over
  • For clusterrole, getting a list of clusterrolebindings initially and using that to check if migration
    is required
• Skipping migration of deployer and builder service accounts which are automatically created
  on OCP (#381)
• Added some default schedule policies (#385)
  • default-migration-policy: Triggers after every minute
  • default-interval-policy: Triggers after every hour. Retains last 10
  • default-daily-policy: Triggers every day at 12:00am. Retains last 7
  • default-weekly-policy: Triggers every Sunday@12:00am. Retains last 4
  • default-monthly-policy: Triggers every 15th day of the month at 12:00am. Retains last 12.
• storkctl now prints N/A in the migration status when volumes or resources are not being migrated (#387)
• Updated the default admin namespace to kube-system (#386)
• Added wait option (-w / --watch) to storkcktl command to activate clusterdomain (#384)
• Added wait option (-w / --watch) to storkcktl command to create migrations (#391)
• Added SyncStatus to the ClusterDomainsStatus CRD indicating whether a cluster domain is in sync with other cluster domains. The status will also be displayed when checking the status with storkctl (#389)
• Added suspend and resume subcommands to storkctl for schedule objects (#390)
• Migration schedules will now be suspended if the local domain is inactive (#382)
• Updated storkctl to print local cluster domain (#382)

Bug Fixes

• Fixed comparing cluster domains to see if an update is required in the status (#368)
• Fixed schedules not getting triggered if the current time is exactly at next trigger time (#371)
• Added missing activation for DeploymentConfig in storkctl activate migrations subcommand (#375)

Docker hub image: openstorage/stork:2.2.3

Release Stork v2.2.2

Improvements

• Added support to migrate the following additional OCP resources (#355)
  • DeploymentConfig
  • ImageStream
  • Route

Container Image Updates

• Updated google cloud sdk to 243.0.0 (#351)
• Switched to python3 from python2.7 (#351)
• Upgraded some python libraries (#351)

Bug Fixes

• Portworx: Fixed memory leak when grpc endpoint was incorrect (#367)

Docker hub image: openstorage/stork:2.2.2

Release Stork v2.2.1

Updates

• Removed google cloud sdk test directories from the container (#349)

Docker hub image: openstorage/stork:2.2.1

Release Stork v2.2.0

New Features

• Added CRD to configure SchedulePolicies (#250)
  • Cluster scoped objects that can be used in Schedule objects
  • Added storkctl subcommands for schedulepolicy
• Added support for scheduling Migrations (#252)
  • A new CRD (MigrationSchedule) has been defined which can be used to run Migrations according to the specified SchedulePolicy
  • Added storkctl subcommands for migrationshotschedule
• Added support for scheduling VolumeSnapshots (#289)
  • A new CRD (VolumeSnapshotSchedule) has been defined which can be used to run VolumeSnapshots according to the specified SchedulePolicy
  • Added storkctl subcommands for volumesnapshotschedule
  • Snapshot schedules can be created automatically on PVC creation by specifying options in the storageclass
  • Examples in #289
• Added support to check and update ClusterDomains (#294)
  • Two new cluster scoped CRDs have been defined
  • ClusterDomainsStatus reports the currently active and inactive domains from the storage driver.
  • ClusterDomainUpdate can be used to update the active and inactive domains
  • Added storkctl subcommands for clusterdomainsstatus and clusterdomainupdate
  • Examples in #294

Improvements

• Added support to run Pre/Post Exec rules during Migration (#238)
  • They can be specified using preExec and postExec fields in the Migration spec
• Added support to specify label selectors during Migration (#272)
  • key: value pairs can be specified using the selectors field to select specific resources to migrate
  • Helpful when you want to migrate only certain apps from a namespace
• Added support to migrate the following additional resources (#313)
  • DaemonSets
  • ServiceAccounts (except default)
  • ClusterRoles (if used in the namespace)
  • ClusterRoleBindings (if used in the namespace)
• Added option to skip pairing of storage when creating ClusterPair (#249)
  • If no storage options are provided only the schedule will be paired
  • This is useful when you have one storage fabric between multiple Kubernetes clusters and only want to migrate resources
• The restore namespace for VolumeGroupSnapshots can now be updated after it was created (#263)
  • This propagates the annotation to all the children snapshots
• Portworx:
  • Added support for authentication for APIs (#324)
  • Added Mode for ClusterPair. Valid options are Default and DisasterRecovery (#327)

Bug Fixes

• Portworx
  • Fixed handling for canceled migrations (#316)
  • Removed compatibility version check for decommissioned nodes (#327)

Docker hub image: openstorage/stork:2.2.0

Release Stork v2.1.2

Improvements

• Portworx: Ability to read service and namespace from environment variables (#307)
  • Service name can be passed in with PX_SERVICE (defaults to portworx-service)
  • Namespace can be passed in with PX_NAMESPACE (defaults to kube-system)