Improved Rancher Managed Network [#2944, #2963, #2901] - Rancher’s overlay networking using IPSec has been refactored with multiple fixes including replacing the current raccoon IKE keying daemon to charon. This should help with overall performance and stability issues found in our older implementation.
- NOTE: This is a fairly large change to our networking and if your network does not upgrade properly due to a timing issues when stopping and starting charon/racoon, please reboot the network agent container on the affected hosts. It’s also possible that during the network agent restart, you may hit issue #1406. If that happens, you will need to reboot the host to correct the networking.
- NOTE2: Please do not upgrade to this version if you are using CoreOS. Please upgrade to v0.56.1.*
UI Enhancements - We have added a few enhancements to the current UI:
- Added a new “System Stack” tab to differentiate between user application stacks vs system stacks (i.e. Convoy and external DNS providers).
- Added a new Help page located on the top navigation that includes integration with Rancher forums, Github projects, and FAQ.
- Improved UI performance and responsiveness.
- Added many UI fixes. Major ones will be listed below.
- Experimental: Added a new optional dark theme to the UI. This is a work in progress as we continue to optimize the UI for a darker theme over the upcoming releases.
Added new health check action strategies [#3076] - The current behavior of health check failure is to immediately destroy and re-create. Starting with this release, Rancher now supports the following strategies:
- destroy - current default behavior
- noop - do nothing and simply provide an indication that the instance is unhealthy
- quorum - allow users to define a quorum size, less than equal to the service scale size, such that Rancher will not attempt to recreate containers until the number of healthy containers is less than the quorum size.
- Rancher Metadata Service Update [#3144] - The metadata service has now been updated to allow queries to /stack/services and /service/containers to return the full objects rather than name only. This is to make queries more efficient without having to query for the containers or services once the name was known. NOTE : We have bumped the current version of metadata service from 2015-07-25 to 2015-12-19 and have also pointed /latest -> 2015-12-19. If you were using /latest prior to this release, please make changes according as the return values have now been changed.
- Added PointHQ DNS [rancher/rancher-catalog#75] Thanks @haswalt!
- Updated Convoy GlusterFS and Convoy NFS [#2796] with improvements to take advantage of the new fixes to managing deletion of volumes. Unfortunately with the major changes that were added, upgrades are not supported from v0.1 to v0.2. A new stack needs to be created.
Known Major Issues
- CoreOS hosts have broken connectivity #3416
Major Bug Fixes since v0.51.0
- Fixed an issue where if the random port allocation for a service fails, the service becomes stuck in a “registering” state forever [#3107]
- Fixed an issue where additional audit logs are added when a service is activated via rancher-compose [#3082]
- Fixed an issue where log_driver and log_opts are not exported correctly to the corresponding YML file [#2388, #3065]
- Fixed a race condition where Route53 records can be deleted [#3263]
- Fixed an issue where if SSL is enabled for a LB, the link generated for the published ip:port does not include “https://” [#3182]
- Fixed an issue where services with the global flag enabled would be destroyed and recreated if their hosts moved from a deactivated to an activated state [#3155]
- Fixed an issue where the AD configuration page is no longer accessible when it has been enabled [#3146]
- Fixed an issue where the accounts page does not properly show the accounts names [#3139]
- Fixed an issue where the metadata service was casting int32/int64 variables to float64 [#3123]
- Fixed an issue where the Rancher catalog does not show parameter defaults for enums and booleans [#3115]
- Hopefully squashed all service LINK display issues [#3090, #1389]
- Fixed an issue where duplicated service names in different stacks results in metadata only returned from 1 stack [#2864, #2091]
- UI now shows the Stack/Service name for LB Services for all services it is forwarding traffic to [#2858]
- UI now shows the Docker version and Kernel version on host [#2766]
- Fixed an issue where AD does not work if a domain already exists for the service (or any) account being added [#2482]
- Fixed an issue where health checks will not work if Rancher selects a host where the network agents are not even running [#2287]
- Added additional validation to port formatting when creating and updating services so it does not cause the entire service to hang [#667]
- Fixed an issue where host labels were being reverted to their previous values if they were added when launching rancher/agent and were using the same key [#3149]
- Stacks will no longer show a “Degraded” status if using services with run-once containers [#2957]
- Added a new “certificate” type for catalog items [#3246]
- Delete volumes now work properly for shared storage [#2796]