Releases: gardener/landscaper
Releases · gardener/landscaper
v0.8.3
[landscaper]
✨ New Features
- [OPERATOR] The defaulting mechanism for the CRD management has been updated to use the default k8s defaulting. (#201, @schrodit)
- Note: the golang configuration struct has slightly changed.
🐛 Bug Fixes
- [USER] Annotations and labels defined on targets are now correctly synced to an installations context. (#200, @schrodit)
- [OPERATOR] The container deployer blueprint now honors the init and wait container defined in its component descriptor. (#206, @schrodit)
- [OPERATOR] A bug has been fixed that caused the a deadlink the
force-reconcile
logic of the execution controller and the deployers. (#203, @schrodit)
🏃 Others
- [DEVELOPER] The component-cli lib has been updated to v0.19.0 (c2de1cb)
v0.8.2
v0.8.1
v0.8.0
[landscaper]
⚠️ Breaking Changes
- [DEVELOPER] The Landscaper specific schemes and decoders have been move from
pkg/kubernetes
topkg/apis
. Please update your imports accordingly. (#151, @schrodit) - [DEVELOPER] Deployers now have to remove timeout annotations (
landscaper.gardener.cloud/reconcile-time
) and updatestatus.observedGeneration
at the beginning of the DeployItem reconcile loop. See the new documentation on expected deployer behaviour for more information. (#143, @Diaphteiros)
✨ New Features
- [OPERATOR] The landscaper now detects if a deploy item is in a non-final state for too long, which is interpreted as the deployer getting stuck while processing it. It will then get the abort operation annotation, giving the deployer the opportunity to abort processing the deploy item gracefully. If that also times out, landscaper will set the deploy item to
Failed
. Have a look at the documentation for further information on the timeouts. (#174, @Diaphteiros) - [OPERATOR] It is now possible to also use labels to select targets for deployers. (#163, @schrodit)
- [OPERATOR] The landscaper now detects if a DeployItem is not reconciled by any deployer within a specified timeframe and puts these DeployItems in
Failed
state. The timeout duration defaults to five minutes and can be set in the landscaper configuration. (#143, @Diaphteiros) - [DEVELOPER] add template function to render image vector (#153, @achimweigel)
🐛 Bug Fixes
- [USER] fixes the passthrough of OCI registry secrets in deployer helm charts (#168, @jschicktanz)
- [DEVELOPER] A bug that prevents the conversion from manifest v1alpha1 to v1alpha2 has been fixed. (#151, @schrodit)
📖 Documentation
- [OPERATOR] Documentation for the deployer controller configuration has been added. (#163, @schrodit)
- In addition the basic target selector is explained in the common deployer docs.
🏃 Others
- [USER] adds option for setting the default repositoryContext in the Landscaper Helm Chart (#160, @jschicktanz)
- [USER] Ingress resources in the documentation and in the examples are updated from v1beta1 to v1. (#156, @MrBatschner)
- [OPERATOR] Integration test for deployer blueprints have been added. (#173, @schrodit)
- [OPERATOR] It is now possible to configure the webhook server image and controller image separately. (#172, @schrodit)
- [OPERATOR] Allows setting Landscaper cluster kubeconfig in all deployer charts and blueprints (#162, @jschicktanz)
- [OPERATOR] The error reporting and error handling of all deployers and the Landscaper controller has been improved by better and more detailed error reporting. (#148, @schrodit)
- [DEVELOPER] A dedicated health helper has been introduced that can be used to determine if a Installation is healhty. (#167, @schrodit)
- [DEVELOPER] Fixes description in GitHub Issue/PR template. (#161, @MrBatschner)
- [DEVELOPER] Elimination of duplication of logs from integration tests. (#159, @MrBatschner)
- [DEVELOPER] The manifest-deployer internally uses v1alpha2 objects exclusively. (#154, @MrBatschner)
- Fixes the wrong version of the Status object, it will be manifest.deployer.landscaper.gardener.cloud/v1alpha2 now.
- Integration test for the manifest-deployer to check if it can handle both, v1alpha1 and v1alpha2 ProviderConfigurations.
v0.7.0
[landscaper]
✨ New Features
- [USER] DeployItem's type is now immutable and cannot be updated. (#144, @schrodit)
- [USER] Helm charts and execution containers can be fetched from private OCI registries with the credentials provided as
Installation.spec.registryPullSecrets
(#130, @loewenstein) - [OPERATOR] The validation webhook and the landscaper controller are now split into different containers and deployment. (#146, @schrodit)
- This is done to ensure isolation of validation and the actual load in the controller.
- Checkout the updated landscaper helm chart.
- [OPERATOR] All deployers can now be installed using the offically created blueprints (#126, @schrodit)
🐛 Bug Fixes
- [USER] Fixed a bug in the Landscaper controller that caused a panic when a imported configmap key was undefined. (#145, @schrodit)
- [USER] A bug has been fixed that caused the provider status of a manifest deploy item to grow infinite when the resource could not be created. (#129, @schrodit)
🏃 Others
v0.6.2
no release notes available
v0.6.1
v0.6.0
[landscaper]
✨ New Features
- [USER] It is now possible to omit the namespace from configmap or secret refs. The namespace is now defaultet to the installation's namespace. (#114, @schrodit)
- [OPERATOR] It is now possible to overwrite deployer configuration when the deployer is part of the landscaper controller. (#121, @schrodit)
- [OPERATOR] The landscaper-controller will now automatically deploy the CRDs for its resources upon startup. (#101, @MrBatschner)
- [OPERATOR] Add a webhook that validates
Installation
,Execution
, andDeployItem
resources. (#64, @Diaphteiros) - [DEVELOPER] The go version has been updated to 1.16.0 (0480f13)
🐛 Bug Fixes
- [USER] A bug has been fixed that causes the landscaper and all deployers to not respect docker-config files. (#122, @schrodit)
- [USER] Several bugs in the container deployer have been fixed that caused the container to not correctly write exports or restore the state. (#121, @schrodit)
- [USER] A bug has been fixed that caused the import validation to stuck when a non-dataref import has been defined on a dependent installation. (#117, @schrodit)
- [USER] The landscaper non-CRD resource jsonschemes are now generated using the Kubernetes OpenAPI-Library. (#109, @schrodit)
- This fixes several issues with the current jsonschemas.
- [OPERATOR] [Manifest Deployer] A bug in the manifest deployer has been fixed that causes it to not properly write its state in case of errors during the apply. (#119, @schrodit)
- [DEVELOPER] A bug has been fixed that prevented usage of the apis module with the landscaper (32af92c)
- [DEPENDENCY] The kubernetes versions are pinned to v1.19.7 to fix a bug in the code-generator in k8s v1.20: kubernetes/kubernetes#98380 (#113, @schrodit)
📖 Documentation
- [USER] A new tutorial and documentation about the usage of external jsonschemas has been added. (#114, @schrodit)
🏃 Others
- [OPERATOR] Expose controller-runtime metrics for landscaper-controller and add oci cache metrics. (#110, @hendrikKahl)
- [OPERATOR] Landscaper's Deployers can now be deployed from their very own Helm charts. (#107, @MrBatschner)
- [DEVELOPER] A integration test for the simple-import and aggregation tutorial has been added. (#117, @schrodit)
- [DEVELOPER] A integration test framework has been added that make it possible to run tests against a cluster with a running landscaper. (#99, @schrodit)
- [DEVELOPER] A integration test has been added that tests the simple nginx-ingress tutorial. (#99, @schrodit)
- [DEPENDENCY] Update controller-runtime dependency to
v0.8.1
(#100, @Diaphteiros) - [DEPENDENCY] The component-cli has been updated to v0.8.0 (51bcd7c)
- [DEPENDENCY] The component-spec has been updated to v0.0.31, which fixes a bug in the component descriptor oci reference generation. (51bcd7c)
v0.5.3
[landscaper]
🐛 Bug Fixes
- [USER] A bug has been fixed that caused the Landscaper to not properly write the state in secrets. (0fc9cd5)
🏃 Others
v0.5.2
no release notes available