Skip to content

Latest commit

 

History

History
705 lines (420 loc) · 19.6 KB

CHANGELOG.md

File metadata and controls

705 lines (420 loc) · 19.6 KB

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog and this project's packages adheres to Semantic Versioning.

3.3.0 - 2024-07-18

Added

  • Added and enabled by default an emptyDir volume under tmp of the deployment for artifact pulls so that we can make the rest of the root file system read only under the security context.
  • Added control over the VPA max allowed resources.

Changed

  • Requests and limits are always set on the deployment.
  • Increased default requests and limits because of the introduction of the emptyDir volume under /tmp.
  • Enabled read only root file system by default on the security context.
  • Update PolicyExceptions to v2 and fallback to v2beta1.

3.2.1 - 2024-03-05

Fixed

  • Use separate rest configs for different Kubernetes clients.

3.2.0 - 2024-02-27

Added

  • Added --service.controller.resyncperiod daemon flag that can be controller via .controller.resyncPeriod Helm value. Controls the duration after which a complete sync with all known runtime objects the controller watches is performed. Defaults to 5m.

3.1.3 - 2024-01-29

Fixed

  • Move pss values under the global property

Changed

  • Use base images from gsoci.azurecr.io

3.1.2 - 2023-12-20

Changed

  • Configure gsoci.azurecr.io as the default container image registry.

3.1.1 - 2023-12-05

Changed

  • Configure gsoci.azurecr.io as the registry to use by default

3.1.0 - 2023-10-19

Changed

  • Force-disable PSP-related resources when global.podSecurityStandards.enforced value is true.

3.0.0 - 2023-10-04

Removed

  • Removed giantswarm.io/monitoring: "true" label from the Service resource. To get metrics chart-operator should be from now on used in conjunction with chart-operator-extensions version v1.1.1 or later to deploy ServiceMonitor resource for it. It was split up as chart-operator is one of the first component to get into a cluster that will deploy most other things, for example Prometheus that will eventually actually deploy the CRD for ServiceMonitor.

2.35.2 - 2023-09-26

Fixed

  • Fixed default values for .proxy and .cluster.proxy values and updated Helm values schema accordingly.

2.35.1 - 2023-09-21

Changed

  • Changed pod taint toleration to only tolerate NotReady for CAPI.

2.35.0 - 2023-05-04

Changed

  • Disable PSPs for k8s 1.25 and newer.

2.34.1 - 2023-05-02

Added

  • Add Kyverno PolicyExceptions for necessary capabilities normally prohibited by PSS policies.

2.34.0 - 2023-02-14

Changed

  • Selecting private Helm client on demand for some operations.

2.33.2 - 2022-12-16

2.33.1 - 2022-12-16

Added

  • New error for values schema validation.

Changed

  • Use transitional errors coming from running Helm in the Chart CR status.

2.33.0 - 2022-11-16

Added

  • Add support to run in private cloud clusters, which cannot provide any working externalDNSIP.

2.32.0 - 2022-11-15

Added

  • Support for running behind a proxy.
    • HTTP_PROXY,HTTPS_PROXY and NO_PROXY are set as environment variables in deployment/chart-operator if defined in values.yaml.
  • Support for using cluster-apps-operator generated cluster.proxy values.

2.31.0 - 2022-10-07

Added

  • Add internal upgrade step on installation for Helm charts marked by annotation.

2.30.0 - 2022-09-23

Added

  • Add suport for timeouts fields in the Chart CR.

Changed

  • Add support for new control-plane label in k8s 1.24.

2.29.0 - 2022-08-12

  • Reconfigure VPA autoscaler to react correctly to pod resource ceilings

2.28.0 - 2022-08-09

Changed

  • Add pre-upgrade helm annotation to giantswarm-critical PriorityClass in order to fix upgrade issues.

2.27.0 - 2022-07-29

Added

  • Ensure the giantswarm-critical PriorityClass is created first on initial installation.

2.26.0 - 2022-07-20

Changed

  • Use 127.0.0.1 as KUBERNETES_SERVICE_HOST when bootstrapMode is enabled.

2.25.0 - 2022-07-04

Changed

  • Tighten pod and container security contexts for PSS restricted policies.
  • Use downward API to set deployment env var KUBERNETES_SERVICE_HOST to status.hostIP.
  • Change initialBootstrapMode configuration value to bootstrapMode.
  • Use private Helm client for installing app-operators from control-plane-test-catalog

Added

  • Allow to set api server pod port when enabling initialBootstrapMode.

2.24.1 - 2022-06-22

Changed

  • Update helmclient to v4.10.1.

2.24.0 - 2022-06-09

Changed

  • Add chart-pull-failed error to differentiate between issues when pulling chart tarball and other problems.

Fixed

  • Fix missing PriorityClass issue.

2.23.0 - 2022-06-06

Changed

  • Always create giantswarm-critical priority class if it does not exist.
  • Add initialBootstrapMode flag to allow deploying CNI as managed apps.

2.22.0 - 2022-05-30

Added

  • Split Helm client into private Helm client for giantswarm-namespaced apps and public Helm client for rest of the apps.

2.21.1 - 2022-05-19

Added

  • Add Helm release failure reason when it is known, and if there is a currently successfully released version

2.21.0 - 2022-04-07

Changed

  • Update helmclient to v4.10.0.

2.20.1 - 2022-03-15

Changed

  • Use apptestctl to install CRDs in integration tests to avoid hitting GitHub rate limits.

Fixed

  • Fix status resource to use Helm release status if it exists.

2.20.0 - 2021-12-15

Changed

  • Update Helm to v3.6.3.
  • Use controller-runtime client to remove CAPI dependency.

Removed

  • Remove unused helm 2 release collector.

2.19.1 - 2021-10-20

Changed

  • Deployment hostNetwork is enabled or not depending on chartOperator.cni.install value.

2.19.0 - 2021-08-13

Removed

  • Remove tillermigration resource now Helm 3 migration is complete.

2.18.1 - 2021-08-05

Changed

  • Increase memory limit for deploying large charts in workload clusters.

2.18.0 - 2021-06-21

Added

  • Add releasemaxhistory resource which ensures we retry at a reduced rate when there are repeated failed upgrades.

Changed

  • Upgrade Helm release when failed even if version or values have not changed to handle situations like failed webhooks where we should retry.

2.17.0 - 2021-06-09

Changed

  • Prepare helm values to configuration management.
  • Update architect-orb to v3.0.0.

Fixed

  • Improve status message when helm release has failed max number of attempts.

2.16.0 - 2021-06-03

Changed

For CAPI clusters:

  • Add tolerations to start on NotReady nodes for installing CNI.
  • Create giantswarm-critical priority class.
  • Use host network to allow installing CNI packaged as an app.

2.15.0 - 2021-05-20

Added

  • Proxy support in helm template.

2.14.0 - 2021-04-30

Changed

  • Cancel the release resource when the manifest object already exists.
  • Cancel the release resource when helm returns an unknown error.

2.13.1 - 2021-04-06

Fixed

  • Updated OperatorKit to v4.3.1 for Kubernetes 1.20 support.

2.13.0 - 2021-03-31

Changed

  • giantswarm-critical PriorityClass only managed when E2E.

2.12.0 - 2021-03-26

Changed

  • Set docker.io as the default registry
  • Pass RESTMapper to helmclient to reduce the number of REST API calls.
  • Updated Helm to v3.5.3.

2.11.0 - 2021-03-19

Added

  • Updating namespace metadata using namespaceConfig in Chart CRs.

2.10.0 - 2021-03-17

Added

  • Pause Chart CR reconciliation when it has chart-operator.giantswarm.io/paused=true annotation.

Changed

  • Deploy giantswarm-critical PriorityClass when it's not found.

2.9.0 - 2021-02-03

Added

  • Use diff key when logging differences between the current and desired release.

Fixed

  • Stop updating Helm release if it has failed the previous 5 attempts.

2.8.0 - 2021-01-27

Added

  • Add support for skip CRD flag when installing Helm releases.

2.7.1 - 2021-01-13

Fixed

  • Only create VPA if autoscaling API group is present.

2.7.0 - 2021-01-07

Added

  • Added last reconciled timestamp as metrics.

2.6.0 - 2020-12-21

Added

  • Print difference between current release and desired release.

Changed

  • Updated Helm to v3.4.2.

2.5.2 - 2020-12-07

Added

  • Add Vertical Pod Autoscaler support.

2.5.1 - 2020-12-01

Fixed

  • Fix comparison of last deployed and revision optional fields in status resource.
  • Set memory limit and reduce requests.

2.5.0 - 2020-11-09

Added

  • Validate the cache in helmclient to avoid state requests when pulling tarballs.
  • Call status webhook with token values.

Fixed

  • Update apiextensions to v3 and replace CAPI with Giant Swarm fork.

2.4.0 - 2020-10-29

Added

  • Call status webhook when webhook annotation is present.

Removed

  • Remove chartmigration resource as migration from chartconfig to chart CRs is complete.

2.3.5 - 2020-10-13

Fixed

  • Stop repeating helm upgrade for the failed helm release.

2.3.4 - 2020-10-01

Added

  • Added release name as a label into the event count metrics.

2.3.3 - 2020-09-29

Fixed

  • Updated Helm to v3.3.4.
  • Updated Kubernetes dependencies to v1.18.9.
  • Update deployment annotation to use checksum instead of helm revision to reduce how often pods are rolled.
  • Increase wait timeout for accessing Kubernetes API from 10s to 120s.

2.3.2 - 2020-09-22

Added

  • Added event count metrics for delete, install, rollback and update of Helm releases.

Fixed

  • Fix structs merging error in helmclient.

Security

  • Updated Helm to v3.3.3.

2.3.1 - 2020-09-04

Added

  • Add monitoring labels.

Changed

  • Add namespace to logging message.

Fixed

  • Remove memory limits from deployment.

2.3.0 - 2020-08-24

Changed

  • Using default DNS policy for control planes.

2.2.1 - 2020-08-19

Changed

  • Fixed the timeout value for the namespace resource.

2.2.0 - 2020-08-19

Changed

  • Creating namespace before helm operations.

2.1.0 - 2020-08-18

Changed

  • Updated Helm to v3.3.0.

2.0.0 - 2020-08-12

Changed

  • Updated backward incompatible Kubernetes dependencies to v1.18.5.
  • Updated Helm to v3.2.4.
  • Fix the rollback in a loop problem.

1.0.7 - 2020-08-05

Changed

  • Rollback the helm release in pending-install, pending-upgrade.

1.0.6 - 2020-07-24

Changed

  • Disable force upgrades since recreating resources is not supported.
  • Graduate Chart CRD to v1.
  • Upgrade to operatorkit 1.2.0.

1.0.5 - 2020-07-15

Changed

  • Enable force upgrades when chart CR annotation is present.

1.0.4 - 2020-07-08

Changed

  • Update MD5 Hash only if chart-operator upgrade the release successfully.
  • Make kubernetes wait timeout configurable when installing and updating releases.
  • Set release revision in CR status.

v1.0.3 2020-06-16

Changed

  • Fixed PodSecurityPolicy compatibility problem.

v1.0.2 2020-06-04

Changed

  • Disabled force-upgrade from helmclient.
  • Canceling the release resource when migration is done yet.

v1.0.1 2020-05-26

Changed

  • Using helmclient v1.0.1 for security fix.
  • Cancel the release resource when the manifest validation failed.

v1.0.0 2020-05-18

Changed

  • Updated to support Helm 3; To keep using Helm 2, please use version 0.X.X.

v0.13.0 2020-04-21

Changed

  • Deploy as a unique app in app collection in control plane clusters.

v0.12.4 2020-04-15

Changed

  • Always set chart CR annotations so update state calculation is accurate.
  • Only update failed Helm releases if the chart values or version has changed.

v0.12.3 2020-04-09

Changed

  • Fix problem pushing chart to default app catalog.

Changed

v0.12.2 2020-04-09

Changed

  • Fix update state calculation and status resource for long running deployments.
  • Handle 503 responses when GitHub Pages is unavailable.
  • Make HTTP client timeout configurable for pulling chart tarballs in AWS China.
  • Switch from dep to go modules.

v0.12.1 2020-03-10

Removed

  • Remove usage of legacy chartconfig CRs in Tiller metrics.

v0.12.0 2020-03-09

Added

  • Add chartmigration resource to allow legacy chartconfig controller to be removed. (#358)

Changed

  • Improve reason field in chart CR status when installing a chart fails. (#359)
  • Use version from chart CR when calculating desired state to reduce number of HTTP requests to pull chart tarballs. (#351)
  • Wait for deleted Helm release before removing finalizer. (#360)
  • Do not wait when installing or updating a Helm release takes over 3 seconds. We check progress in the next reconciliation loop. (#362)

Removed

  • Remove legacy chartconfig controller. (#365)

Added

  • Separate network policy.

Added

  • Separate podsecuritypolicy.
  • Security context in deployment spec with non-root user.