Releases: projectsveltos/addon-controller
v0.57.2
v0.57.1
🚀 Features
1. Support for CEL Rules
Introduced CEL (Common Expression Language) rule support across multiple resource types. These rules enhance dynamic evaluation and validation capabilities, and can now be defined for the following instances:
- Classifier
- EventSource
- HealthCheck
- SveltosCluster liveness and readiness checks
- Techsupport
2. Configurable Token Duration for Token Renewal
Added an optional tokenDuration field to the TokenRequestRenewalOption struct, allowing fine-grained control over token validity:
If tokenDuration is set, it defines how long the token remains valid.
If not set, the system falls back to renewTokenRequestInterval.
This ensures tokens can be renewed proactively, reducing risk during brief network outages.
3. Event Generation on Feature Deployment
The system now emits events when features are successfully deployed to a cluster via ClusterSummary. These events provide visibility into the deployment process and include details about the feature and target cluster.
4. EventTrigger destinationCluster Field
Introduced a new field, destinationCluster, in the EventTrigger resource. This field, a corev1.ObjectReference, can be expressed as a template. It addresses scenarios where an event in one cluster (e.g., a management cluster) necessitates deploying resources to a different, specific cluster.
If defined, Sveltos will first instantiate destinationCluster (using both Resource and source cluster data), then set the ClusterProfile.Spec.ClusterRefs.
v0.56.0
🐛 Bug Fixes
-
Helm Chart Upgrade/Downgrade Errors Not Reported
Fixed an issue where errors during Helm chart upgrades or downgrades were being ignored. Sveltos would incorrectly mark these operations as successful even when failures occurred.
PR #1224 -
Incorrect Cleanup of Stale Resources on Deployment Failure
Addressed a bug where Sveltos would incorrectly delete existing Kubernetes resources when deploying an update using PolicyRefs or KustomizationRefs. If a deployment failed partway (e.g., due to invalid data), previously deployed resources not explicitly redeployed in the failed attempt were being treated as stale and removed.
PR #1220 -
Deployment Failure with Long Cluster/Profile Names
Resolved an issue in continuousWithDriftDetectionMode where deployments would fail if the combination of cluster name and profile name exceeded 63 characters.
PR #1230 -
Dashboard Crash Due to File Descriptor Limit
Fixed a bug in the dashboard that caused it to crash with a "too many open files" error.
v0.55.0
🚀 Features
- Values Merge Support: Helm chart values specified via Values and ValuesFrom are now properly merged, following the defined order (first all Values, then ValuesFrom). This is powered by mergo. mergo is used for this goal
👉 PR
🐛 Bug Fixes
- Drift Detection Optimization (Agentless Mode):
1. Improved performance by limiting the cache to only load ResourceSummary instances for the relevant cluster.
2. Fixed a slow startup issue caused by inefficient drift detection initialization.
3. Resolved a bug where stale ResourceSummary instances remained after cluster deletion. 👉 PR #1207 - Advanced Dashboard Dependencies:
1. 👉 PR #90 - Addon controller:
1. Improved Helm Failure Handling: Enhanced how Sveltos responds to Helm installation failures for better reliability. 👉 PR #1206
2. Conflict Handover Fix: Resolved an issue where Profile B couldn't take over management of a Helm chart after it was removed from Profile A, despite the conflict being cleared. 👉 PR
v0.54.0
🚀 Features
- Templated Resource Identifiers: The namespace and name fields of resources referenced in PolicyRefs, TemplateResourceRefs, and ValuesFrom can now be defined using templates based on any field of the target cluster—enabling highly flexible and reusable configurations.
- Event-Driven Templating: Resources referenced by EventTrigger now support templated namespace and name values, with full access to both the triggering cluster and the resource that generated the event.
- Optional Resources Support: Referenced resources can be marked as optional. If not found, Sveltos will continue execution without raising an error.
🐛 Bug Fixes
- Kubeconfig Refresh in Agentless Mode: In sveltos-agent mode, the agent now correctly reloads the kubeconfig after token renewal to ensure uninterrupted access to managed clusters.
- Resilient Service Discovery: Improved handling of Kubernetes discovery API errors. Sveltos now gracefully processes partial failures from extension APIs (e.g., metrics), continuing with the available resources.
v0.53.0
🚀 Features
- The LabelFilter Operation has been extended to include Has (checks for the presence of a key) and DoesNotHave (checks for the absence of a key) in addition to the existing Equal and Different operations.
🐛 Bug Fixes
v0.52.3
🐛 Bug Fixes
- Agentless Mode Event Processing: Resolved an issue where event processing in the management cluster stopped after upgrading from a version prior to v0.52 in agentless mode. Fix
- Addon-Controller Logging: Fixed a harmless but verbose error message related to drift detection in the management cluster logs reported by the addon-controller. Fix
v0.52.2
v0.52.1
🐛 Bug Fixes
- Bug Fix: Avoid OwnerReferences on Managed Cluster Resources. Sveltos now uses annotations instead of OwnerReferences for resources deployed on managed clusters. This prevents failures that occurred when deploying resources to clusters where Sveltos is also running, as cluster-wide resources like ClusterProfile cannot be OwnerReferences for namespaced resources.
v0.52.0
🚀 Features
- In agentless mode, Sveltos leaves no footprint on managed clusters. Instead, all resources necessary for event detection, configuration drift analysis, and reloads are created and managed directly within the management cluster, ensuring the same level of performance.