10 KiB
| title | description | keywords | redirect_from | |
|---|---|---|---|---|
| UCP 2.1 release notes | Release notes for Docker Universal Control Plane. Learn more about the changes introduced in the latest versions. | Docker, UCP, release notes |
|
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, to upgrade your installation to the latest release.
Version 2.1.2
(29 Mar 2017)
Known issues
There is known issue in UCP 2.1 where upgrading from UCP 1.1.z can cause swarm to leave worker nodes in a pending state with the message:
[Pending] Completing node registration
There are two workarounds for rectifying this issue:
- When upgrading from UCP 1.1.z, first upgrade to UCP 2.0.z, and then to UCP 2.1.z. This will prevent the issue from happening, and is the recommended upgrade path.
- If you have already upgraded from UCP 1.1.z directly to UCP 2.1.z, you can fix the issue by restarting the ucp-swarm-manager container on each of your UCP controller nodes.
This issue will be fixed in UCP 2.1.3.
Bug fixes
-
Core
ucp-reconcileservice now correctly brings upucp-kvcontainer if it has stopped or become unreachable- Fixed known issue in which users are unable to log into UCP UI after upgrading from UCP 2.1.0 to 2.1.1 because the parameter for maximum concurrent users was incorrectly defaulted to '0'
- Fixed an issue where the UCP manager becomes unresponsive and requires a restart
if
docker psordocker infocalls to engine take a long time for a response - HTTP Routing Mesh now correctly provides httplog for debug logging of services
docker node ls -fnow correctly filters when run against a UCP clusterdocker inspect taskno longer returns errors when run against a UCP cluster- UCP now correctly reports progress when loading an image from CLI
-
docker/ucp image
- UCP support dumps now include Docker Engine daemon logs
- Host address IPs are now automatically added to SANs during install
- UCP now reports its version number in the CLI after being installed
-
UI/UX
- Deploying Compose-based applications in the GUI now works correctly when Docker Content Trust "Run Only Signed Images" is turned on
- Fixed an issue where UI temporarily showed more tasks for a service than actually existed
- Fixed an issue in which metrics incorrectly displayed
0%in the UI
Version 2.1.1
(14 Mar 2017)
Known issues
If you are currently running UCP 2.1.0 and previously customized the sessions lifetime parameter in the Authentication settings UI, upgrading to UCP 2.1.1 may cause users to not be able to log into UCP and DTR. This is caused by a faulty default value which sets maximum concurrent user sessions to zero.
You can either wait for UCP 2.1.2 to be released so that the problem is automatically fixed, or upgrade to 2.1.1, and use the following steps to fix the problem.
Start by getting the current configuration for user sessions by running:
curl -u admin "https://$UCP_HOST/enzi/v0/config/sessions"
The command will prompt for the admin user's password and then return
the current sessions config which should look something like:
{
"lifetimeHours": 72,
"renewalThresholdHours": 24,
"perUserLimit": 0
}
If perUserLimit is set to 0, you need to set it to a value between 1 and 100.
The recommended value is 5. You should also customize the command below with
the lifetimeHours and perUserLimit values returned by the first command.
curl -u admin "https://$UCP_HOST/enzi/v0/config/sessions" \
-X PUT \
-H 'Content-Type: application/json' \
-d '{"lifetimeHours": 72, "renewalThresholdHours": 24, "perUserLimit": 5}'
You'll now be able to log into UCP and DTR.
New features
- Core
- Administrators can now configure the frequency with which UCP polls metrics.
Use
docker service update --env-add METRICS_SCRAPE_INTERVAL=10m ucp-agent, and the frequency can be in s/m/h/d. - Administrators can now configure the frequency with which UCP gathers disk usage data.
Use
docker service update --env-add METRICS_DISK_USAGE_INTERVAL=12h ucp-agent, and the frequency can be in s/m/h/d. - Support for syncing users and teams from multiple LDAP servers/domains
(e.g. a separate server to use for
dc=domain2,dc=example,dc=com) - Support for limiting the number of maximum concurrent login sessions any user may have
- Administrators can now configure the frequency with which UCP polls metrics.
Use
Bug fixes
- Core
- Fixed an issue in which UCP manager would panic and be unable to return the right system status after the cluster became unhealthy
ucp-hrmcontainer now provides debug logs throughstdout- HTTP Routing Mesh now checks to ensure an ingress port is not already in use by UCP or DTR before becoming active
- Fixed an issue in which UCP did not use swarm-mode node IDs, preventing usage of node constraints and other features when using cloned VMs as UCP nodes
- Fixed an issue in which certain Docker API 1.26 commands were not correctly supported
- Disk usage metrics no longer display 0% when using devicemapper filesystem
- Disk usage metrics are now collected every 2 hours by default, and can be tunned
- Fixed an issue causing Content Trust enforcement to ignore an optional
tagfor/images/create, causing some signed content to not run correctly - LDAP sync logs now take up less disk space on manager nodes
- UCP support dumps are now correctly compressed to take up less disk space, and provide information on HTTP Routing Mesh and metrics
- docker/ucp image
- UCP install now correctly fails and presents an error when trying to
specify
host-addressto an existing swarm-mode cluster - Clarified upgrade message to make it clear that the upgrade command now works at once for the entire cluster rather than needing to be run on every node
- UCP install now correctly fails and presents an error when trying to
specify
- UI/UX
- UI now displays a warning if there is significant latency or network issues in communications between UCP manager nodes
- UI no longer incorrectly displays 'No Services' while still loading the Services tab
- UI no longer displays errors when global tasks are removed due to node constraints
- UI now displays a warning when underlying engines in the swarm-mode cluster are running different versions
- UI now displays an error when 'Load Image' command fails
- 'KV Store Timeout' option now displays correct units (milliseconds)
- Dashboard now correctly displays errors when metrics are unavailable
- The DTR deployment page now validates if a DTR replica ID is valid or not
Version 2.1.0
(9 Feb 2017)
This version of UCP extends the functionality provided by CS Docker Engine 1.13. Before installing or upgrading this version, you need to install CS Docker Engine 1.13 in the nodes that you plan to manage with UCP.
New features
-
Core
- Support for managing secrets (e.g. sensitive information such as passwords or private keys) and using them when deploying services. You can store secrets securely on the cluster and configure who has access to them, all without having to give users access to the sensitive information directly
- Support for Compose yml 3.1 to deploy stacks of services, networks, volumes, and secrets.
- HTTP Routing Mesh now generally available. It now supports HTTPS passthrough where the TLS termination is performed by your services, Service Name Indication (SNI) extension of TLS, multiple networks for app isolation, and Sticky Sessions
- Granular label-based access control for secrets and volumes (NOTE: unlike other resources controlled via label-based access control, a volume without a label is accessible by all UCP users with Restricted Control or higher default permissions)
-
UI/UX
- You can now view and manage application stacks directly from the UI
- You can now view cluster and node level resource usage metrics
- When updating a service, the UI now shows more information about the service status
- Rolling update for services now have
failure-actionwhich you can use to - Several improvements to service lifecycle management specify rollback, pausing, or continuing if the update fails for a task
- LDAP synching has more configuration options for extra flexibility
- UCP now warns when the cluster has nodes with different Docker Engine versions
- The HTTP routing mesh settings page now lists all services using the routing mesh, with details on parameters and health status
- Admins can now view team membership in a user's details screen
- You can now customize session timeouts in the authentication settings page
- Can now mount
tmpfsor existing local volumes to a service when deploying services from the UI - Added more tooltips to guide users on the above features
Bug fixes
-
Core
- HTTP routing mesh can now be enabled or reconfigured when UCP is configured to only run images signed by specific teams
- Fixed an error in which
_pingcalls were causing multiple TCP connections to open up on the cluster - Fixed an issue in which UCP install occasionally failed with the error "failed to change temp password"
- Fixed an issue where multiple rapid updates of HTTP Routing Mesh configuration would not register correctly
- Demoting a manager while in HA configuration no longer causes the
ucp-auth-apicontainer to provide errors
-
UI/UX
- When creating a user, pressing enter on keyboard no longer causes problems
- Fixed assorted icon and text visibility glitches
- Installing DTR no longer fails when "Enable scheduling on UCP controllers and DTR nodes" is unchecked.
- Publishing a port to both TCP and UDP in a service via UI now works correctly
Known issues
The docker stats command is sometimes wrongly reporting high CPU usage.
Use the top command to confirm the real CPU usage of your node.
Learn more.
Version compatibility
UCP 2.1 requires minimum versions of the following Docker components:
- Docker Engine 1.13.0
- Docker Remote API 1.25
- Compose 1.9