Skip to content

Latest commit

 

History

History
508 lines (354 loc) · 33.6 KB

CHANGELOG.md

File metadata and controls

508 lines (354 loc) · 33.6 KB
Raw

Changelog

v0.11.0 - 2020-03-05

Changelog since v0.10.0. For changelog since v0.11.0-beta.3, please check the release notes

Attention Needed

  • Kubernetes 1.14 clusters are not supported as of this release because 1.14 isn't supported by the upstream anymore
    • It remains possible and is advisable to upgrade 1.14 clusters to 1.15
    • Currently, it also remains possible to provision 1.14 clusters, but that can be dropped at any time and it'll not be fixed if it stops working
  • As of this release, it is not possible to upgrade 1.13 clusters to 1.14
    • Please use an older version of KubeOne in the case you need to upgrade 1.13 clusters
  • The AWS Terraform configuration has been refactored a in backward-incompatible way (#729)
    • Terraform now handles setting up subnets
    • All resources are tagged ensuring all cluster features offered by AWS CCM are supported
    • The security of the setup has been increased
    • Access to nodes and the Kubernetes API is now going over a bastion host
    • The aws-private configuration has been removed
    • Check out the new Terraform configuration for more details

Added

  • Add support for Kubernetes 1.17
    • Fix cluster upgrade failures when upgrading from 1.16 to 1.17 (#764)
  • Add support for ARM64 clusters (#783)
  • Add ability to deploy Kubernetes manifests on the provisioning time (KubeOne Addons) (#782)
  • Add the kubeone status command which checks the health of the cluster, API server and etcd (#734)
  • Add support for NodeLocalDNSCache (#704)
  • Add ability to divert access to the Kubernetes API over SSH tunnel (#714)
  • Add support for sourcing proxy settings from Terraform output (#698)
  • Persist configured proxy in the system package managers (#749)

Changed

General

  • [Breaking] The AWS Terraform configuration has been refactored (#729)
  • The KubeOneCluster manifests is now parsed strictly (#802)
  • The leader instance can be defined declarative using API (#790)
  • Make vSphere Cloud Controller Manager read credentials from a Secret instead from cloud-config (#724)

Bug fixes

  • Fix CentOS cluster provisioning (#770)
  • Fix AWS shared credentials file handling (#806)
  • Fix credentials handling if .cloudProvider.Name is none (#696)
  • Fix upgrades not determining hostname correctly causing upgrades to fail (#708)
  • Fix kubeone reset failing to reset the cluster (#727)
  • Fix configure-cloud-routes bug for AWS causing kube-controller-manager to log warnings (#725)
  • Disable validation of replicas count in the workers definition (#775)
  • Proxy settings defined in the config have precedence over those defined in Terraform (#760)

Updates

  • Update machine-controller to v1.9.0 (#774)
  • Update Canal CNI to v3.10 (#718)
  • Update metrics-server to v0.3.6 (#720)
  • Update DigitalOcean Cloud Controller Manager to v0.1.21 (#722)
  • Update Hetzner Cloud Controller Manager to v1.5.0 (#726)

Removed

  • Remove ability to upgrade 1.13 clusters to 1.14 (#764)
  • Removed FlexVolume support from the Canal CNI (#756)

Docs

  • GCE clusters must be configured as Regional to work properly (#732)

v0.11.0-beta.3 - 2019-12-20

Attention Needed

  • Kubernetes 1.14 clusters are not supported as of this release because 1.14 isn't supported by the upstream anymore
    • It remains possible and is advisable to upgrade 1.14 clusters to 1.15
    • Currently, it also remains possible to provision 1.14 clusters, but that can be dropped at any time and it'll not be fixed if it stops working
  • As of this release, it is not possible to upgrade 1.13 clusters to 1.14
    • Please use an older version of KubeOne in the case you need to upgrade 1.13 clusters

Added

  • Add support for Kubernetes 1.17
    • Fix cluster upgrade failures when upgrading from 1.16 to 1.17 (#764)

Removed

  • Remove ability to upgrade 1.13 clusters to 1.14 (#764)

v0.11.0-beta.2 - 2019-12-12

Changed

Bug Fixes

  • Proxy settings defined in the config have precedence over those defined in Terraform (#760)

v0.11.0-beta.1 - 2019-12-10

Added

  • Persist configured proxy in the system package managers (#749)

Changed

Bug fixes

  • Fix kubeone status reporting wrong etcd status when the hostname is not provided by Terraform or config (#753)

Removed

  • Removed FlexVolume support from the Canal CNI (#756)

v0.11.0-beta.0 - 2019-11-28

Attention Needed

  • The AWS Terraform configuration has been refactored a in backward-incompatible way (#729)
    • Terraform now handles setting up subnets
    • All resources are tagged ensuring all cluster features offered by AWS CCM are supported
    • The security of the setup has been increased
    • Access to nodes and the Kubernetes API is now going over a bastion host
    • The aws-private configuration has been removed
    • Check out the new Terraform configuration for more details

Added

  • Add the kubeone status command which checks the health of the cluster, API server and etcd (#734)
  • Add support for NodeLocalDNSCache (#704)
  • Add ability to divert access to the Kubernetes API over SSH tunnel (#714)
  • Add support for sourcing proxy settings from Terraform output (#698)

Changed

General

  • [Breaking] The AWS Terraform configuration has been refactored (#729)
  • Make vSphere Cloud Controller Manager read credentials from a Secret instead from cloud-config (#724)

Bug fixes

  • Fix credentials handling if .cloudProvider.Name is none (#696)
  • Fix upgrades not determining hostname correctly causing upgrades to fail (#708)
  • Fix kubeone reset failing to reset the cluster (#727)
  • Fix configure-cloud-routes bug for AWS causing kube-controller-manager to log warnings (#725)

Updates

  • Update machine-controller to v1.8.0 (#736)
  • Update Canal CNI to v3.10 (#718)
  • Update metrics-server to v0.3.6 (#720)
  • Update DigitalOcean Cloud Controller Manager to v0.1.21 (#722)
  • Update Hetzner Cloud Controller Manager to v1.5.0 (#726)

Docs

  • GCE clusters must be configured as Regional to work properly (#732)

v0.10.0 - 2019-10-09

Attention Needed

  • Kubernetes 1.13 clusters are not supported as of this release because 1.13 isn't supported by the upstream anymore
    • It remains possible to upgrade 1.13 clusters to 1.14 and is strongly advised
    • Currently, it also remains possible to provision 1.13 clusters, but that can be dropped at any time and it'll not be fixed if it stops working
  • KubeOne now uses Go modules! 🎉 (#550)
    • This should not introduce any breaking changes
    • If you're using go get to obtain KubeOne, you have to enable support for Go modules by setting the GO111MODULE environment variable to on
    • You can obtain KubeOne v0.10.0 using the following go get command: go get github.com/kubermatic/kubeone@v0.10.0

Known Issues

  • kubectl scale is not working with kubectl v1.15 to due an upstream issue. Please upgrade to kubectl v1.16 if you want to use the kubectl scale command.

Added

  • Add support for Kubernetes 1.16
    • Fix cluster upgrade failures when upgrading from 1.15 to 1.16 (#670)
    • Update default admission controllers for 1.16 (#667)
  • Add support for sourcing credentials and cloud-config file from a YAML file (#623, #641)
  • Add the StaticAuditLog feature used to configure the audit log backend (#631)
  • Add SystemPackages API field used to control configuration of repositories and packages (#670)
  • Add support for managing clusters over a bastion host (#567, #651)
  • Add support for specifying OpenStack Tenant ID using the OS_TENANT_ID environment variable (#551)
  • Add ability to configure static networking for worker nodes (#606)
  • Add taints on control plane nodes by default (#564)
  • Add ability to apply labels on the Node object using the .workers.providerSpec.Labels field (#677)
  • Add ability to apply taints on the worker nodes using the .workers.providerSpec.Taints field (#678)
  • Add an optional rootDiskSizeGB field to the worker spec for OpenStack (#549)
  • Add an optional nodeVolumeAttachLimit field to the worker spec for OpenStack (#572)
  • Add an optional TrustDevicePath field to the worker spec for OpenStack (#686)
  • Add optional BillingCycle and Tags fields to the worker spec for Packet (#686)
  • Add ability to use AWS spot instances for worker nodes using the isSpotInstance field (#686)
  • Add support for Hetzner Private Networking (#596)
  • Add ShortNames and AdditionalPrinterColumns for Cluster-API CRDs (#689)
  • Add an example KubeOne Ansible playbook (#576)

Changed

Bug Fixes

  • Fix kubeone install and kubeone upgrade generating v1beta1 instead of v1beta2 kubeadm configuration file for 1.15 and 1.16 clusters (#670)
  • Fix cluster provisioning failures when the DynamicAuditLog feature is enabled (#630)
  • Fix kubeone reset not retrying deleting worker nodes on error (#639)
  • Fix kubeone reset not skipping deleting worker nodes if the machine-controller CRDs are not deployed (#683)
  • Fix Terraform integration not respecting multiple workerset definitions from output.tf (#568)
  • Fix kubeone install failing if Terraform output is not provided (#574)
  • Flannel CNI is forced use an internal network if it's available (#598)

Updates

Proxy

  • Write proxy configuration to the /etc/environment file (#687, #688)
  • Fix proxy configuration file (/etc/kubeone/proxy-env) generation (#650)
  • Fix machine-controller and machine-controller-webhook deployments not receiving the proxy configuration (#657)

Examples

  • Fix GCE provisioning failure when using a longer cluster name with the example Terraform script (#607)

Removed

  • Remove TemplateNetName field from the vSphere workers spec (#624)
  • Remove the old KubeOne configuration API (#626)
    • This should not affect you unless you're using the pre-v0.6.0 configuration API manually
    • The kubeone config migrate command is still available, but might be deleted at any time

v0.10.0-alpha.3 - 2019-08-22

Changed

  • Update machine-controller to v1.5.3 (#633)

v0.10.0-alpha.2 - 2019-08-21

Changed

  • Fix cluster provisioning failures when DynamicAuditLog feature is enabled (#630)
  • Update machine-controller to v1.5.2 (#624)

Removed

  • Remove TemplateNetName field from the vSphere workers spec (#624)
  • Remove the old KubeOne configuration API (#626)

v0.10.0-alpha.1 - 2019-08-16

Added

  • Add ability to configure static networking for worker nodes (#606)

Changed

v0.10.0-alpha.0 - 2019-07-17

Attention Needed

  • KubeOne now uses Go modules! 🎉 (#550)
    • This should not introduce any breaking change
    • If you're using go get to obtain KubeOne, you may have to enable support for Go modules by setting the GO111MODULE environment variable to on

Added

  • Add support for SSH over a bastion host (#567)
  • Add an optional rootDiskSizeGB field to the worker spec for OpenStack (#549)
  • Add an optional nodeVolumeAttachLimit field to the worker spec for OpenStack (#572)
  • Add support for specifying OpenStack Tenant ID using the OS_TENANT_ID environment variable (#551)
  • Add an example KubeOne Ansible playbook (#576)

Changed

  • Fix Terraform integration not respecting multiple workerset definitions from output.tf (#568)
  • Fix install failing if Terraform output is not provided (#574)
  • Update machine-controller to v1.4.2 (#572)
  • Control plane nodes are now tainted by default (#564)

v0.9.2 - 2019-07-04

Changed

  • Fix the CNI plugin URL for cluster upgrades on CoreOS (#554)
  • Fix kubelet binary upgrade failure on CoreOS because of binary lock (#556)

v0.9.1 - 2019-07-03

Changed

  • Fix .ClusterNetwork.PodSubnet not being respected when using the Weave-Net CNI plugin (#540)
  • Fix kubeadm preflight check failure (IsDockerSystemdCheck) on Ubuntu and CoreOS by making Docker use systemd cgroups driver (#536, #541)
  • Fix kubeadm preflight check failure on CentOS due to kubelet service not being enabled (#541)

v0.9.0 - 2019-07-02

Action Required

  • The Terraform integration now requires Terraform v0.12+
    • Please see the official Upgrading to Terraform v0.12 document to find out how to update your Terraform scripts for v0.12
    • The example Terraform scripts coming with KubeOne are already updated for v0.12
    • KubeOne is not able to parse output of terraform output generated with Terraform v0.11 or older anymore
    • The Terraform output template (output.tf) has been changed and KubeOne is not able to parse the old template anymore. You can check the output template used by example Terraform scripts as a reference

Added

  • Add support for Kubernetes 1.15 (#486)
  • Add support for Microsoft Azure (#469)
  • Add kubeone completion command for generating the shell completion scripts for bash and zsh (#484)
  • Add kubeone document command for generating man pages and KubeOne documentation (#484)
  • Add support for reading Terraform output directly from the directory (#495)
  • Add missing fields to the workers specification API (#499)

Changed

  • [BREAKING] KubeOne Terraform integration now uses Terraform v0.12+ (#466)
  • [BREAKING] Change Terraform output template to conform with the KubeOneCluster API (#503)
  • Fix Docker not starting properly for some providers on Ubuntu (#512)
  • Fix kubeone reset failing if a MachineSet or Machine object has been already deleted and include more details in the error messages (#508)
  • Fix ability to read Terraform output from the standard input (stdin) (#479)
  • Update machine-controller to v1.3.0 (#499)
  • Update DigitalOcean Cloud Controller Manager to v0.1.15 (#516)
  • Use Docker 18.09.7 when provisioning new clusters (#517)
  • Configure proxy for kubelet on control plane nodes if proxy settings are provided (#496)
  • Configure proxy on worker nodes if proxy settings are provided (#490)
  • Make GoBetween Load balancer configuration script work on all operating systems and fix minor bugs (#494)

v0.8.0 - 2019-05-30

Added

  • Add support for VMware vSphere (#428)

v0.7.0 - 2019-05-28

Added

  • Add WeaveNet as an additional CNI plugin (#432)
  • Add the --remove-binaries flag to the kubeone reset command that removes Kubernetes binaries (#450)

Changed

  • Fix kubeone reset failing if no MachineDeployment or Machine object exist (#450)
  • Update machine-controller to v1.1.8 (#454)

v0.6.2 - 2019-05-13

Changed

  • Fix a missing JSON tag on the Name field, so specifying name in lowercase in the manifest works as expected (#439)
  • Fix failing to disable SELinux on CentOS if it's already disabled (#443)
  • Fix setting permissions on the remote KubeOne directory when the $USER environment variable isn't set (#443)

v0.6.1 - 2019-05-09

Changed

  • Provide the --kubelet-preferred-address-types flag to metrics-server, so it works on all providers (#424)

v0.6.0 - 2019-05-08

We're excited to announce that as of this release KubeOne is in beta! We have the new backwards compatibility policy going in effect as of this release.

Check out the documentation for this release to find out how to get started with KubeOne.

Action Required

  • This release introduces the new KubeOneCluster API. The new API is supposed to the improve user experience and bring many new possibilities, like API versioning.
    • Old KubeOne configuration manifests will not work as of this release!
    • To continue using KubeOne, you need to migrate your existing manifests to the new KubeOneCluster API. Follow the migration guidelines to find out how to migrate.

Added

  • [BREAKING] Implement and migrate to the KubeOneCluster API (#343, #353, #360, #379, #390)
  • Implement the config migrate command for migrating old configuration manifests to KubeOneCluster manifests (#408)
  • Implement the config print command for printing an example KubeOneCluster manifest (#412, #415)
  • Enable scaling subresource for MachineDeployments (#334)
  • Deploy metrics-server by default (#338, #351)
  • Deploy external cloud controller manager for DigitalOcean, Hetzner, and Packet (#364)
  • Implement Terraform integration for Hetzner (#331)
  • Add support for OpenIDConnect configuration (#344)
  • Add support for Packet provider (#384)
  • Patch CoreDNS deployment to work with external cloud controller manager taints (#362)

Changed

  • Fix a typo in vSphere CloudProviderName (#339)
  • Expose ssh_username variable in the Terraform output (#350)
  • Pass -external-cloud-provider flag to machine-controller when external cloud provider is enabled (#361)
  • Update machine-controller to v1.1.5 (#378)
  • Don't wait for machine-controller if it's not deployed (#392)
  • Deploy instance with a load balancer on OpenStack. General improvements to OpenStack support (#401)
  • Parse all parameters from Terraform output for DigitalOcean (#370)

v0.5.0 - 2019-04-03

Added

  • Add support for Kubernetes 1.14
  • Add support for upgrading from Kubernetes 1.13 to 1.14
  • Add support for Google Compute Engine (#307, #317)
  • Update machine-controller when upgrading the cluster (#304)
  • Add timeout after upgrading each node to let nodes to settle down (#316, #319)

Changed

  • Deploy machine-controller v1.1.2 on the new clusters (#317)
  • Creating MachineDeployments and upgrading nodes tasks are repeated three times on the failure (#328)
  • Allow upgrading to the same Kubernetes version (#315)
  • Allow the custom VPC to be used with the example AWS Terraform scripts and switch to the T3 instances (#306)

v0.4.0 - 2019-03-21

Action Required

  • In #264 the environment variable names were changed to match names used by Terraform. In order for KubeOne to correctly fetch credentials you need to use the following variables:
    • DIGITALOCEAN_TOKEN instead of DO_TOKEN
    • OS_USERNAME instead of OS_USER_NAME
    • HCLOUD_TOKEN instead of HZ_TOKEN
  • The cloud-config file is now required for OpenStack clusters. Validation will fail if the cloud-config is not provided. See the OpenStack quickstart for details how the cloud-config file should look like.
  • Ark integration is removed from KubeOne in #265. You need to remove the backups section from the existing configuration files. If you wish to deploy Ark on new clusters you have to do it manually.

Added

  • Add support for enabling DynamicAuditing (#261)

Changed

  • Deploy machine-controller v1.0.7 on the new clusters (#247)
  • Automatically download the Kubeconfing file after install (#248)
  • Default --destory-workers to true for the kubeone reset command (#252)
  • Improve OpenStack Terraform scripts (#253)
  • The cloud-config file for OpenStack is required (#253)
  • The environment variable names are changed to match names used by Terraform (#264)

Removed

  • Option for deploying Ark is removed (#265)

v0.3.0 - 2019-03-08

Action Required

  • If you're using provider.Name external to configure control plane nodes to work with external CCM you need to:
    • Set provider.Name to name of the cloud provider you're using or to none (see config.yaml.dist for supported values),
    • Set provider.External to true.
    • Note: using external CCM is not supported and is currently not working as expected.
  • If you're using provider.Name none:
    • If you're deploying machine-controller you need to set MachineController.Provider to name of the cloud provider you're using for worker nodes,
    • Otherwise you need to set MachineController.Deploy to false, so validation passes.

Added

  • Add support for cluster upgrades (#206, #211, #214)
  • Add support for enabling PodSecurityPolicy (#218)
  • Add kubeone version command (#221)
  • Add initial support for OpenStack worker nodes (#209)

Changed

  • Fix none provider bugs and ensure correct behavior (#227)
  • External Cloud Controller Manager is now enabled by setting Provider.External to true instead of setting Provider.Name to external (#230, #237)
  • Set correct Internal address on nodes with 2+ network interfaces (#240)
  • Deploy machine-controller on uninitialized nodes (#241)

v0.2.0 - 2019-02-15

Added

  • Add support for workers creation on DigitalOcean (#153)
  • Add support for CentOS (#154)
  • Add support for external cloud controller managers (#161)
  • Add Terraform scripts for OpenStack (#170)
  • Add support for configuring proxy for Docker daemon, apt-get/yum and curl (#182)

Changed

  • AWS credentials can be obtained from the profile file (#156)
  • Fix Etcd init on clusters with more than one network interface (#160)
  • Allow provider.Name to be specified for providers without external Cloud Controller Manager (#161)
  • Refactor KubeOne CLI (#177)
    • Global flags are parsed correctly regardless of their placement in the command
    • --verbose and --tfjson are global flags
    • The main package is moved to the root of project (github.com/kubermatic/kubeone)
  • Deploy machine-controller v1.0.4 instead of v0.10.0 to fix CVE-2019-5736 (#191)
  • Deploy Docker 18.09.2 on Ubuntu to fix CVE-2019-5736 (#193)

Removed

  • Remove support for Kubernetes 1.12 (#184)
  • Remove API fields related to Etcd (#194)