Add Beta1 release notes (docker#340)

Author: Jim Galasyn

datacenter/ucp/3.0/guides/release-notes.md

@@ -1,5 +1,5 @@
 ---
-title: UCP 3.0 release notes
+title: UCP 3.0 Beta1 release notes
 description: Release notes for Docker Universal Control Plane. Learn more about the
   changes introduced in the latest versions.
 keywords: UCP, release notes
@@ -9,3 +9,248 @@ Here you can learn about new features, bug fixes, breaking changes, and
 known issues for the latest UCP version.
 You can then use [the upgrade instructions](admin/install/upgrade.md), to
 upgrade your installation to the latest release.
+
+## Version 3.0 Beta1
+
+(11 December 2017)
+
+**New features**
+
+* UCP now supports Kubernetes as an orchestrator, in addition to the existing Swarmkit 
+and "Classic" Swarm orchestrators. Kubernetes system components are installed automatically on all manager and worker nodes in the cluster. Kubernetes in UCP is not yet supported on Windows or IBM Z nodes. 
+* Worker nodes can be allocated to run only Swarm workloads, only Kubernetes workloads, or mixed workloads. Mixed workloads are not recommended for use in a production environment due to potential resource contention issues across orchestrators. Manager nodes are by default Mixed in order to support Swarm and Kubernetes system components.
+* Hostname Routing Mesh (HRM) has been upgraded to use Interlock 2 backend for layer 7 routing. This adds increased performance, stability, and new features including SSL Termination, Contextual Path-based Routing, Websocket Support, and Canary Application Instance deployments. Existing HRM labels (and newly added labels with the old format) will automatically migrate to the new format. It is recommended to use the new format for new applications in order to take advantage of the added features.
+
+**Known issues**
+
+* UCP 3.0 Beta1 has been tested on Ubuntu 16.04 and RHEL 7.3
+* Installation on SLES 12 and Ubuntu 14.04 is not currently possible because of an iptables 
+incompatibility
+* UCP 3.0 requires more resources to run than UCP 2.2 and is unlikely to work 
+correctly on nodes with less than 4GB of total memory.
+* Security hardening of Kubernetes managed by Docker EE is not fully complete. 
+Only use this release for testing and validation in controlled environments with 
+trusted users. In-progress hardening also means that some Kubernetes conformance 
+tests do not currently pass. Specifically, apps that rely on the default Kubernetes 
+service account may not work because of access restrictions
+* The product versions reported by components in the beta release are preliminary 
+and will change for the GA final release
+* Interlock, the successor to the HTTP Routing Mesh (HRM), has not yet been fully 
+scale-tested or optimized and the default settings in the beta release are not 
+optimized for production use
+* HRM supports configuring a catch-all fallback service for requests that do not 
+match any routing directive. Interlock does not support this feature
+* When changing the mode for a worker node between Kubernetes, Swarm and Mixed, 
+when going back to "Swarm" some Kubernetes tasks may not be evicted from the node
+* A Kubernetes bug causes workloads published using `NodePort` to only be accessible 
+on the particular nodes that are running pods for the workload
+* Problems with state-reconciliation may cause nodes that are changed from managers 
+to workers to get into a state where they're incorrectly running Kubernetes master 
+components
+* Installing on systems with SELinux in enforcing mode currently fails because of 
+a Calico installation problem
+* Deleting Kubernetes Pods may leave pods in "Terminating" state with no way to 
+delete them
+* Removing a node from Docker Swarm may not remove the node from the Kubernetes 
+node set
+* When promoting nodes from worker to master, not all required labels and taints 
+are correctly applied. This can cause cluster failure if master nodes are lost #9842
+Reconciliation may fail on RHEL and CentOS because of problems with mounting the 
+`/var` folder
+
+## Version 2.2.4
+
+(2 November 2017)
+
+**News**
+
+* Docker Universal Control Plane now supports running managers on IBM Z on RHEL, SLES and Ubuntu. Previously, only workers were supported on IBM Z.
+
+**Bug fixes**
+
+* Core
+  * `ucp-etcd` system images are now hidden. Previously, these system images were erroneously displayed in the images list (#9913).
+  * `disable_usageinfo` will now disable usage metrics. A regression caused this setting to not be respected (#9469).
+  * UCP now outputs "Initializing..." log messages during setup so that administrators can establish that setup or install has begun (#9336).
+  * Windows worker promotion is now blocked. Previously, Windows workers could be promoted using the CLI, which would fail (#9341).
+  * Loading gzipped images with the Docker CLI is now supported. This would previously cause a panic (#6367).
+  * Permissions are now checked when filtering nodes by container. Previously, permissions were not considered (#9239).
+  * An LDAP sync is now triggered as soon as an LDAP user is lazy-provisioned. Previously, lazy-provisioned users would not immediately be added to teams and orgs (#5674).
+
+* UI/UX
+  * License page now shows all capabilities. Previously it was not clear if a license supported Docker image scanning or not (#9946).
+  * Additional translations added for internationalization (#9559).
+  * UI for adding users to teams simplified (#8399, #9404).
+  * The grant list can now sorted and pagination in the grants view has been improved. The grants view previously had glitches on systems with many grants (#8874, #9203).
+  * Fixed problem where UI would hang when pulling images (#9380).
+  * "Max failure ratio" and "Failure action" re-introduced in service definitions. These settings were not available in UCP 2.2, but were available in previous UCP versions (#9186).
+  * Collection labels are no longer applied to UCP system services. UCP previously auto-applied labels, which was confusing (#10030)
+
+**Known issues**
+
+ * Docker currently has limitations related to overlay networking and services using VIP-based endpoints. These limitations apply to use of the HTTP Routing Mesh (HRM). HRM users should familiarize themselves with these limitations. In particular, HRM may encounter virtual IP exhaustion (as evidenced by `failed to allocate network IP for task` Docker log messages). If this happens, and if the HRM service is restarted or rescheduled for any reason, HRM may fail to resume operation automatically. See the Docker EE 17.06-ee5 release notes for details.
+ * The Swarm admin UI for UCP versions 2.2.0 and later contain a bug. If used with Docker Engine version 17.06.2-ee5 or earlier, attempting to update "Task History Limit", "Heartbeat Period" and "Node Certificate Expiry" settings using the UI will cause the cluster to crash on next restart. Using UCP 2.2.X and Docker Engine 17.06-ee6 and later, updating these settings will fail (but not cause the cluster to crash). Users are encouraged to update to Docker Engine version 17.06.2-ee6 and later, and to use the Docker CLI (instead of the UCP UI) to update these settings. Rotating join tokens works with any combination of Docker Engine and UCP versions. Docker Engine versions 17.03 and earlier (which use UCP version 2.1 and earlier) are not affected by this problem.
+
+## Version 2.2.3
+
+(13 September 2017)
+
+**Bug fixes**
+
+* Core
+  * Node list will no longer show duplicated worker node entries.
+  * Volume mount options are no longer dropped when creating volumes.
+  * `docker stack deploy` with secrets specified in docker-compose file now works.
+* UI/UX
+  * Upgrade button is now greyed out and deacticated after initiating upgrade.
+  * If an error is encountered while creating a service, the UI no longer freezes.
+  * Upgrade notification fixed to have working link.
+  * "Default Role For All Private Collections" can now be updated. Updating this
+  role in the UI previously had no effect.
+  * Added notification to UI to show that an upgrade is in progress.
+  * Client bundle can now be downloaded with Safari browser.
+  * Windows nodes are no longer displayed in the DTR install UI.
+  * DTR settings state in UCP is now preserved when switching tabs. Previously,
+  un-saved state was lost when switching tabs.
+  * Fixed problem where first manager node may have IP address `0.0.0.0`,
+  causing dashboard to not update.
+  * UI for adding Windows nodes improved to include full join instructions.
+  * Node Task UI fixed. Displaying tasks for a node previously did not work.
+  * LDAP settings UI improved. Sync interval setting is now validated, a
+  never-ending update spinner been fixed and it's UI action sequencing bugs have
+  been fixed so that it's now possible to disable LDAP.
+  * Uploading Docker images in the UI now has better error messages and improved
+  validation.
+  * Containers removed in UI are now force-removed. Previously removing
+  containers would fail.
+  * DTR install instructions `--ucp-url` parameter fixed to have valid value.
+  * Deleting multiple users in succession fixed. Previously, an error would
+  result when deleting more than one user at a time.
+  * Added validation when adding DTR URL in UCP admin settings.
+  * Left-nav now shows resource counts, addressing an UI regression from UCP 2.1.
+
+**Known issues**
+
+ * Upgrading heterogeneous swarms from CLI may fail because x86 images are used
+ instead of the correct image for the worker architecture.
+ * Agent container log is empty even though it's running correctly.
+ * Rapid UI settings updates may cause unintended settings changes for logging
+ settings and other admin settings.
+ * Attempting to load an (unsupported) `tar.gz` image results in a poor error
+ message.
+ * Searching for images in the UCP images UI doesn't work.
+ * Removing a stack may leave orphaned volumes.
+ * Storage metrics are not available for Windows.
+ * You can't create a bridge network from the web UI. As a workaround use
+ `<node-name>/<network-name>`.
+
+
+## version 2.2.2
+
+(30 August 2017)
+
+**Bug fixes**
+
+* Core
+  * Fixed an issue that caused timeouts during install, preventing UCP 2.2.1 from
+  being released.
+  * Fixed a number of issues in which access control labels and roles could not
+  be upgraded to their new format, when upgrading UCP.
+  [Learn more](https://success.docker.com/KBase/Auth_system_migration_errors).
+  * Fixed an issue that caused an upgrade with multiple manager nodes to fail
+  with RethinkDB startup errors.
+  * Fixed an issue that caused upgrades to fail due to UCP being unable to
+  remove and replace older UCP containers.
+  * Fixed an issue in which upgrade timed out due to lack of available disk space.
+  * Fixed an issue in which rescheduling of containers not belonging in services
+  could fail due to a request for a duplicate IP address.
+  * DTR containers are no longer omitted from `docker ps` commands.
+* UI/UX
+  * Fixed known issue from 2.2.0 where config changes (including LDAP/AD) take
+  an extended period to update after making changes in the UI settings.
+  * Fixed an issue where the `/apidocs` url redirected to the login page.
+  * Fixed an issue in which the UI does not redirect to a bad URL immediately
+  after an upgrade.
+  * Config and API docs now show the correct LDAP sync cron schedule format.
+* docker/ucp image
+  * Support dump now contains information about access control migrations.
+  * The `ucp-auth-store` and `ucp-auth-api` containers now report health checks.
+
+**Known issues**
+
+* When deploying compose files that use secrets, the secret definition must
+include `external: true`, otherwise the deployment fails with the error
+`unable to inspect secret`.
+
+## Version 2.2.0
+
+(16 August 2017)
+
+**New features**
+
+* The role-based access control system has been overhauled for additional
+granularity and customization. Admins now define access control through Grants,
+a 1:1:1 mapping of a Subject, a Role, and a Collection:
+  * Subject: A user, team, or organization.
+  * Role: A set of permissions. In addition to the existing predefined roles,
+  admins can now create custom roles with their choice of permissions taken
+  from the full Docker API.
+  * Collection: A group of containers or container-based resources (e.g. volumes,
+  networks, secrets, etc.). Collections have a hierarchical directory-like structure
+  and replace the old access control labels from the previous system (though they
+  still use labels in the CLI).
+  * [Read the documentation](access-control/index.md#transition-from-ucp-21-access-control)
+   for more information and examples of the new system and how your old access
+   control settings are migrated during an upgrade.
+* UCP now provides access control for nodes, where an admin can enforce
+physical isolation between users on different nodes in the cluster. This means two
+different teams can only view and deploy on the nodes to which they have access.
+This is only available with an EE Advanced license.
+* Enhancements to the user management system:
+  * UCP now supports the user concept of organizations, which are groups of teams.
+  * Users can now specify a default collection which automatically applies
+  access control labels to all CLI deploy commands when no label is specified by
+  the user.
+* Support for UCP workers running Windows Server 2016, and the ability to deploy
+Windows-based containerized applications on the cluster.
+  * [Read the documentation](admin/configure/join-windows-worker-nodes/index.md)
+  for instructions on how to join Windows nodes, and current limitations when
+  deploying Windows applications.
+* Support for UCP workers running on IBM Z systems with RHEL 7.3, Ubuntu 16.04,
+and SLES 12.
+* UCP now provides a public, stable API for cluster configuration and access control,
+and the API is fully interactive within the UCP UI.
+* Support for using services with macvlan networks and configuring network scope in UI.
+* The UCP UI has been redesigned for ease-of-use and data management:
+  * Redesigned dashboard with time-series historical graphs for usage metrics.
+  * Compact layout to more easily view resource information at a glance.
+  * Detail panels for resources no longer slide out and cover the main panel.
+  * Filtering mechanism to display related items (e.g. resources in a collection or stack).
+
+**Known issues**
+
+* UI issues:
+  * Cannot currently remove nodes using UCP UI. Workaround is to remove from CLI
+  instead.
+  * Search does not function correctly for images.
+  * Cannot view label constraints from a collection's details pages. Workaround
+  is to view by editing the collection.
+  * Certain config changes to UCP make take several minutes to update after making
+  changes in the UI. In particular this affects LDAP/AD configuration changes.
+  * Turning `LDAP Enabled` from "Yes" to "No" disables the save button. Workaround
+  is to do a page refresh which completes the configuration change.
+  * Removing stacks from the UI may cause certain resources to not be deleted,
+  including networks or volumes. Workaround is to delete the resources directly.
+  * When you create a network and check 'Enable hostname based routing', the web
+  UI doesn't apply the HRM labels to the network. As a workaround,
+  [create the network using the CLI](https://docs.docker.com/datacenter/ucp/2.2/guides/user/services/use-domain-names-to-access-services/#service-labels).
+  * The web UI does not currently persist changes to session timeout settings.
+  As a workaround you can update the settings from the CLI, by [adapting these instructions for the
+  session timeout](https://docs.docker.com/datacenter/ucp/2.2/guides/admin/configure/external-auth/enable-ldap-config-file/).
+* docker/ucp
+  * The `support` command does not currently produce a valid support dump. As a
+  workaround you can download a support dumps from the web UI.
+* Windows issues
+  * Disk related metrics do not display for Windows worker nodes.
+  * If upgrading from an existing deployment, ensure that HRM is using a non-encrypted
+  network prior to attaching Windows services.