Changelog for Kubermatic 2.10

v2.10.2

Dashboard:

  • Updated Dashboard to v1.2.2
    • Missing parameters for OIDC providers have been added.
    • containerRuntimeVersion and kernelVersion are now displayed on NodeDeployment detail page
    • Fixed changing default OpenStack image on Operating System change
    • The OIDC provider URL is now configurable via oidc_provider_url variable.

Misc:

  • Insecure Kubernetes versions v1.13.6 and v1.14.2 have been disabled.

v2.10.1

Bugfix:

  • A bug that caused errors on very big addon manifests was fixed
  • fixed kube-state-metrics in user-clusters not being scraped
  • Updated the machine-controller to fix the wrong CentOS image for AWS instances
  • vSphere VMs are cleaned up on ISO failure.

Misc:

  • updated Prometheus to v2.9.2
  • Draining of nodes now times out after 2h
  • the API stops creating an initial node deployment for new cluster for KubeAdm providers.
  • More details are shown when using kubectl get machine/machineset/machinedeployment
  • Pod AntiAffinity and PDBs were added to the Kubermatic control plane components and the monitoring stack to spread them out if possible and reduce the chance of unavailability
  • Support for Kubernetes 1.11.10 was added

v2.10.0

Features

Kubermatic core

  • ACTION REQUIRED: Change config option Values.kubermatic.rbac to Values.kubermatic.masterController
  • Add user cluster controller manager: It is deployed within the cluster namespace in the seed and takes care of reconciling all resources that are inside the user cluster
  • Add feature gate to enable etcd corruption check
  • Remove Kubernetes 1.10 as officially supported version from Kubermatic as it is EOL
  • Add short names to the ClusterAPI CRDs to allow using kubectl get md for machinedeployments, kubectl get ms for machinesets and kubectl get ma to get machines
  • Update canal to v2.6.12, Kubernetes Dashboard to v1.10.1 and replace kube-dns with CoreDNS 1.3.1
  • Update Vertical Pod Autoscaler to v0.5
  • Stop using the name “kubermatic” for cloud provider resources visible to end users
  • Deploy the node-exporter by default as an addon into user clusters to provide Grafana dashboards for user cluster resource usage
  • Make the default AMI’s for AWS instances configurable via the datacenters.yaml
  • Stop deploying Vertical Pod Autoscaler by default
  • Create initial node deployments inside the same API call as the cluster to fix spurious issues where the creation didn’t happen
  • Create events on the object for errors when reconciling MachineDeployments and MachineSets
  • Filter out not valid VM types for azure provider
  • Mark cluster upgrades as restricted if kubelet version is incompatible
  • Enable automatic detection of the OpenStack BlockStorage API version within the cloud config
  • Add the ContainerLinuxUpdateOperator to all clusters that use ContainerLinux nodes
  • Enable the configuration of trust-device-path cloud config property of Openstack clusters via datacenters.yaml.
  • Set AntiAffinity for pods to prevent situations where the API servers of all clusters got scheduled on a single node
  • Set resource requests and limits for all addons
  • Add Kubernetes v1.14.1 to the list of supported versions
  • Ensure reservation of a small amount of resources on each node for the Kubelet and system services
  • Update etcd to v3.3.12
  • Update the metrics-server to v0.3.2
  • Update the user cluster Prometheus to v2.9.1
  • Enable the scaling of MachineDeployments and MachineSets via kubectl scale

Dashboard

  • Change the color scheme
  • Enable the editing of the project name in UI
  • Refine nodes and node deployments statuses
  • Redesign DigitalOcean sizes and OpenStack flavors option pickers
  • Smooth operation on bad network connection by changes in asset caching
  • Add a flag allowing to change the default number of nodes created with clusters
  • Allow setting of OpenStack tags for instances via UI
  • Allow node deployment naming
  • Enable adding multiple owners to a project via UI
  • Allow specifying kubelet version for Node Deployments
  • Display events related to the Nodes in the Node Deployment details view
  • Show project owners in project list view
  • Add possibility to assign labels to nodes
  • Update AWS instance types
  • Add option to include custom links into the application
  • Remove AWS instance types t3.nano & t3.micro as they are too small to schedule any workload on them
  • Redesign the application sidebar

Logging & Monitoring stack

  • Update fluent-bit to 1.0.6
  • Add elasticsearch-exporter to logging stack to improve monitoring
  • Create alerts for cert-manager created certificates about to expire
  • Add blackbox-exporter chart
  • Update Elasticsearch to 6.6.2
  • Add Grafana dashboards for kubelet metrics
  • Update Prometheus to 2.8.1 (Alertmanager 0.16.2) and Grafana to 6.1.3
  • Allow configuration of Alertmanager PVC size
  • Add lifecycle hooks to the Elasticsearch StatefulSet to make starting/stopping more graceful
  • Stop logging pod annotations in Elasticsearch
  • Improve Prometheus backups in high traffic environments
  • Fix VolumeSnapshotLocations for Ark configuration
  • Stop exposting node-exporter on all host interfaces
  • Improve Kibana usability by auto-provisioning index patterns
  • Add configurable Prometheus backup timeout to accomodate larger seed clusters

Other

  • ACTION REQUIRED: update from Ark 0.10 to Velero 0.11
  • Replace handwritten go tcp proxy with Envoy within the nodeport-proxy
  • Update cert-manager to 0.7.0, Dex to 2.15.0, Minio to RELEASE.2019-04-09T01-22-30Z
  • Update nginx-ingress-controller to 0.24.1
  • Allow scheduling Helm charts using affinities, node selectors and tolerations for more stable clusters
  • Allow defining configurable resource constraints for Helm charts
  • Improve Helm charts metadata to make Helm-based workflows easier and aid in cluster updates
  • Enable configuration of dex keys expirations in helm chart
  • Update the nodeport-proxy Envoy to v1.10

Bugfixes

  • Fix invalid variable caching in Grafana dashboards
  • Fix reload behaviour of OpenStack setting fields
  • Fix a bug with the missing version in the footer
  • Fix display number of replicas if the field is empty (0 replicas)