Breaking Changes from 1.x
- DC/OS Kubernetes
2.1.1-1.12.5requires DC/OS 1.12.
- DC/OS Kubernetes
2.1.1-1.12.5introduces breaking changes to the way the package works and is deployed. Therefore, it is not possible to upgrade an existing installation of DC/OS Kubernetes to
- Before installing
kubernetespackage must be installed and running.
- It is no longer possible to install DC/OS Kubernetes on DC/OS Enterprise without specifying a service account and a service account secret with adequate permissions.
- Package options have been renamed and re-organized.
control_plane_reserved_resourcescontains the combined resources from previous
control_plane_diskwere moved to under this group.
- Kubernetes v1.12.5
- Kubernetes Dashboard v1.10.1
- dcos-commons v0.55.0
- Docker v18.09.1
- CoreDNS v1.2.6
- Calico v3.2.4
etcd. When set, etcd will check all incoming peer requests from the cluster for valid client certificates signed by the supplied CA.
- Add the new flag
cluster updatecommand to force the update of the cluster configuration.
- Move the validation of the service configuration to the Mesosphere Kubernetes Engine.
cluster restorecommands. The AWS session token can now be used as part of the AWS credentials.
- Support relative paths in
cluster kubeconfigcommand, e.g.
- Documentation section on how to upgrade the
- Enable the selection of the desired region where to deploy the Kubernetes cluster.
- Increase the number of retries an etcd task will perform during installation to resolve its own DNS name. This should prevent etcd tasks from getting stuck in a retry loop on larger clusters.
- Fixed a bug that might cause
kube-node-publictasks to freeze in the
STARTEDstate, causing installations or upgrades to stop indefinitely.
- Fixed a bug that could forever fail to run public Kubernetes node tasks.
- Fixed a bug affecting node decommission that could cause Kubernetes apps temporary downtime.
- Fixed a bug affecting use of private Docker registries.
- Fixed a bug that might cause segfault when running
dcos kubernetes cluster kubeconfig
- Fixed a bug that might cause pods that have resource limits crash on RHEL based systems. The issue is related to Linux kmem accounting turned-on by default by runc. We now turn-off kmem accounting on RHEL-based systems, and on these systems alone. No user intervention is needed, however all of Kubernetes cluster tasks will be replaced, which may cause some downtime.
- Add an Overview page explaining in detail what changed since the 1.x series of releases.
- Add a CLI page detailing the new Mesosphere Kubernetes Engine CLI.
Advanced Installationpage merging its content into Customizing your Installation.
- Add a Private Docker Registry page explaining how to configure it.
Known issues and limitations are listed in the Limitations page.