Releases: projectsveltos/access-manager
v0.21.1
🐛 Bug Fixes
- addon-controller not reacting to new cluster matching an existing profile bug
- addon-controller not cleaning up ClusterConfiguration for a deleted cluster which used to match a Profile bug
- resource deployed in drift-detection mode has spurious redeployment due to an issue detecting false configuration drift bug
- addon-controller logs reported an issue collecting ResourceSummary and ReloaderReport. That was a non issue as error was happening when ResourceSummary or the ReloaderSummary CRD were not deployed in a cluster bug
v0.21.0
🚀 Features:
- Introducing Profile CRD: similar to ClusterProfile, Profile has a cluster selector and a list of add-ons and applications to deploy in matching clusters. Unlike ClusterProfile, Profile is a namespaced resource and can only match clusters in its same namespace and reference resources in its same namespace. This makes Profile useful for managing multiple tenants in the management cluster, where each tenant has its own namespace.
- EventSource and Classifier now have the capability to analyze resources of different kinds simultaneously. This feature is enabled by the ResourceSelectors field, which allows you to select resources from various GroupVersionKinds and filter them based on various criteria, including: Namespace, Labels and Custom Lua functions for more sophisticated filtering. The optional AggregatedSelection field provides an additional layer of filtering for the resources selected using ResourceSelectors. This field accommodates a Lua function that further narrows down the subset of resources that have already been identified through ResourceSelectors. The function receives the array of resources selected by ResourceSelectors and processes it to eliminate or retain specific resources based on the specified criteria. If AggregatedSelection is not specified, all resources selected by ResourceSelectors will be considered. This feature enables more intricate filtering and selection operations on the resources, allowing for comprehensive analyses that consider all resources simultaneously. This capability is beneficial for more sophisticated tasks, such as identifying resources that are interconnected or share similar properties.
further select a subset of the resources that have already been selected using the ResourceSelector field. - Discord notification: on top of Slack and Webex, notifications can be sent to Discord channel as well
- EventSource and Classifier CRDs have changed
- EventBasedAddOn CRD has been renamed to EventTrigger
- When using template, both Cluster and SveltosCluster can be referenced with keyword "Cluster"
- OpenAPI validations was removed (compliance validations can only be expressed using Lua)
v0.20.0
🚀 Features:
- ClusterProfile dependencies: a clusterProfile instance can now depend on other ClusterProfile instances. If ClusterProfile A depends on ClusterProfile B, for every matching cluster, add-ons and applications listed in ClusterProfile A won't be provisioned till all add-ons and applications listed in ClusterProfile B are provisioned.
🐛 Bug Fixed:
- When ClusterProfile is instructed to create resources in the management cluster, those resources were incorrectly considered stale and removed when new managed clusters started matching. This behaviour is now fixed by this commit.
v0.19.0
🚀 Features:
Introducing cluster sharding: when Sveltos is managing hundreds of clusters, it is advised to use cluster sharding to spread the load between multiple instances of Sveltos controllers. To enable horizontal scaling, add annotation on cluster sharding.projectsveltos.io/key
.
The Sveltos shard-controller
will detect new cluster shards and start new set of Sveltos controllers to handle all clusters in a given shard.
v0.18.2
v0.18.1
v0.18.0
🚀 Features:
- Sveltos can be configured to start agents (
sveltos-agent
anddrift-detection-manager
) in the management cluster. Default behaviour is to start those agents in each managed cluster - SveltosCluster has new field
KubeconfigName
which allows overriding the default sveltos convention which expected a valid kubeconfig to be hosted in a secret with the pattern${sveltosClusterName}-sveltos-kubeconfig.
v0.17.0
🚀 Features:
Rolling Update Strategy for ClusterProfile
A ClusterProfile might match more than one cluster. When adding or modifying a ClusterProfile, it is helpful to:
- Incrementally add the new configuration to a few clusters at a time.
- Validate health before declaring deployment successful.
This pull request introduces two new ClusterProfile Spec fields:
-
MaxUpdate
: Indicates the maximum number of clusters that can be updated concurrently.
Value can be an absolute number (e.g., 5) or a percentage of desired pods (e.g., 10%). Defaults to 100%.- Example: When this field is set to 30%, when the list of add-ons/applications in ClusterProfile
changes, only 30% of matching clusters will be updated in parallel. Only when updates in those clusters
succeed will other matching clusters be updated.
- Example: When this field is set to 30%, when the list of add-ons/applications in ClusterProfile
-
ValidateHealths
: A slice of health validation expressed using the Lua language.- For instance, when deploying Helm charts, it is possible to instruct Sveltos to check deployment
health (number of active replicas) before declaring the Helm chart deployment successful.
- For instance, when deploying Helm charts, it is possible to instruct Sveltos to check deployment
Benefits of a Rolling Update Strategy
A rolling update strategy allows you to update your clusters gradually, minimizing downtime and risk.
By updating a few clusters at a time, you can identify and resolve any issues before rolling out the
update to all of your clusters. Additionally, you can use the ValidateHealths field to ensure that
your clusters are healthy before declaring the update successful.
How to Use the Rolling Update Strategy
To use the rolling update strategy, simply set the MaxUpdate field in your ClusterProfile Spec to
the desired number of clusters to update concurrently.
You can also use the ValidateHealths field to specify any health validation checks that you want to perform.
For example, the following ClusterProfile Spec would update a maximum of 30% of matching clusters concurrently
and would check that the number of active replicas is greater than zero before declaring the update successful:
apiVersion: config.projectsveltos.io/v1alpha1
kind: ClusterProfile
metadata:
name: kyverno
spec:
clusterSelector: env=fv
syncMode: Continuous
maxUpdate: 30%
helmCharts:
- repositoryURL: https://kyverno.github.io/kyverno/
repositoryName: kyverno
chartName: kyverno/kyverno
chartVersion: v3.0.1
releaseName: kyverno-latest
releaseNamespace: kyverno
helmChartAction: Install
values: |
admissionController:
replicas: 1
validateHealths:
- name: deployment-health
featureID: Helm
group: "apps"
version: "v1"
kind: "Deployment"
namespace: kyverno
script: |
function evaluate()
hs = {}
hs.healthy = false
hs.message = "available replicas not matching requested replicas"
if obj.status ~= nil then
if obj.status.availableReplicas ~= nil then
if obj.status.availableReplicas == obj.spec.replicas then
hs.healthy = true
end
end
end
return hs
end
🐛 Bug Fixed:
Leftover ClusterConfigurations and ClusterSummaries after CAPI cluster deletion (more details projectsveltos/addon-controller#325)
v0.16.0
🚀 Features:
Projectsveltos can now manage also the management cluster.
Management cluster can be registered to be managed by Sveltos either as any other cluster or using sveltosctl register mgmt-cluster
v0.15.3
🚀 Features:
- Make management referenced resource name template: ClusterProfile can references resources in the management cluster which need to be fetched and whose values are then used to configure resources to be deployed in managed clusters. Those resources' namespaces/names can be expressed as template and reference Cluster namaspace/name
🐛 Bug Fixed:
- Reconcile ConfigMap/Secret when annotation change