NewYour coding agent can read the release notes before it upgrades.Set up the MCP server →
Go modules, #2430 by repository stars
Latest stable
v1.17.0
Last release 14 days ago
08 Nov 2026
Ships on a steady schedule
a new release about every 8 days
Nearly every release is documented
notes for 59 of the last 60 stable releases
Nothing withdrawn
no release was ever pulled
3 years old
1229 releases · first in 2023
Drag across the ruler to compare, or tap a release to open it.
Nothing published for this version
This release introduces significant enhancements to the Sveltos Dashboard, providing deeper visibility into event-driven workflows, and brings critica
This release introduces significant enhancements to the Sveltos Dashboard, providing deeper visibility into event-driven workflows, and brings critical performance optimizations to EventReports and Pull-Mode cluster management. We’ve also enhanced the Sveltos MCP Server to simplify debugging of the Sveltos event framework.
The Dashboard now offers a comprehensive view of your event-driven automation.
The Sveltos MCP (Model Context Protocol) Server has been enhanced to assist in debugging the Sveltos event framework. This tool provides a structured way to interact with and inspect the internal state of the event engine.
Reference: mcp-server PR #33
We have optimized how EventReports are handled. Previously, every collected report was reprocessed. Now, Sveltos performs a delta check and only reprocesses EventReports that have changed since the last collection, significantly reducing CPU cycles in the management cluster.
Improved handling for managed clusters in Pull Mode when the Sveltos Agent stops reporting a "keep-alive":
Fault Detection: If an agent is non-responsive, Sveltos will skip that cluster and report a specific error: agent is not sending keep alive.
Auto-Recovery: As soon as the agent resumes communication, Sveltos automatically reprocesses all resources that were in a faulty state during the downtime.
Drift-Detection: Fixed a critical issue where the drift-detection-manager failed to detect expired Kubeconfigs when running as an agent in a managed cluster.#363).
Event Manager: Enhanced reliability for event-driven resource distribution (#436).
Event Manager: Improved cleanup and state management within the event manager (#427).
Addon Controller: Various stability fixes and internal logic improvements across PRs #1572, #1568, #1563, and #1555.
Reloader: Fixed an issue where triggered rollouts failed for DaemonSets and StatefulSets when the Reloader knob was enabled.
Your coding agent can read these notes before it upgrades. Set up the MCP server →
Resolved an issue where the drift-detection-manager was being unnecessarily upgraded in managed clusters even when drift detection was not configured
Resolved an issue where the drift-detection-manager was being unnecessarily upgraded in managed clusters even when drift detection was not configured (i.e., when a Profile did not use SyncMode: ContinuousWithDriftDetection). The controller now correctly checks for the presence of the ResourceSummary CRD before initiating the drift-detection-manager upgrade.
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Introduced the optional PreHealthCheckDeployment field to the _ClusterPromotion_ resource. This allows users to deploy and reconcile specific resource
Introduced the optional PreHealthCheckDeployment field to the ClusterPromotion resource. This allows users to deploy and reconcile specific resources—such as validation Jobs or temporary services—after the configured delay but before post-delay health checks are executed. This ensures the environment is fully prepared for validation.
Sveltos now supports granular, per-cluster configuration for its agents. Users can apply unique configuration patches to individual managed clusters using annotations to reference a specific ConfigMap. This is ideal for scenarios requiring cluster-specific overrides, such as using regional private container registries.
Introduced the sveltosctl redeploy cluster command, allowing users to forcefully trigger a re-application of all Sveltos-managed add-ons and resources on a target cluster.
Intelligent Execution: Instead of a simple reset, it performs a dependency-aware reconciliation. It uses a topological sort based on the dependsOn field in ClusterSummary resources to ensure prerequisites are reconciled before dependent resources.
Benefit: Enables a full configuration refresh or rolling update without requiring any changes to the ClusterProfile or Profile specifications.
Introduced the TakeOwnership field for Helm chart deployments. This allows Sveltos to manage and adopt resources that already exist in a cluster, even if they lack the standard Helm ownership annotations (meta.helm.sh/release-name, etc.).
Default: false (standard Helm safety checks remain active unless explicitly enabled). PR
Fixed ActiveWindow Reconciliation: Resolved an issue where changes to the ActiveWindow settings in a SveltosCluster resource would not always trigger a correct recalculation of the NextPause and NextUnpause times. This fix ensures that any modification to the ActiveWindow specification forces a re-evaluation of the scheduling. Additionally, the feature now explicitly reads and stores ActiveWindow times using UTC to prevent timezone-related errors in scheduling calculations. PR
Fixed Drift Exclusion Reconciliation: Fixed a bug where Sveltos would not react to changes in the DriftExclusions field within a ClusterProfile or Profile. This issue prevented Sveltos from correctly updating the set of resources excluded from drift detection when the configuration was modified. PR
Fixed Stuck Add-on Deletion in Pull Mode: Corrected an issue where, in pull mode, add-on deletion would get stuck if the governing Profile or ClusterProfile was deleted while the resources were still being deployed. This occurred because the local controller attempted deletion while the remote agent's action was still set to deploy. The fix ensures the remote ConfigurationGroup is cleaned up, allowing the deletion to complete. PR
Action Required: To upgrade to this Sveltos release, users leveraging ClusterAPI integration must ensure their management clusters are using v1beta2 ClusterAPI CRDs.
Due to the fix for DriftExclusions tracking, the internal hash calculation now includes the driftExclusion field.
Impact: Upon upgrading to this version, any Profile or ClusterProfile that contains populated driftExclusion fields will undergo a one-time automatic redeployment.
ClusterAPI: v1.12.0 Go: v1.25.5 Helm: v3.19.4
Filter deployed resources by namespace, name, kind
A new Custom Resource Definition (ClusterPromotion) is introduced to enable declarative, multi-stage, progressive rollout of a single configuration (p
A new Custom Resource Definition (ClusterPromotion) is introduced to enable declarative, multi-stage, progressive rollout of a single configuration (profileSpec) across multiple sets of clusters. This is ideal for promoting configurations from environments like staging to production in a controlled, sequential manner.
It manages an ordered series of stages, each with its own cluster selector.
It supports automated promotion to the next stage after the previous stage's configuration is successfully deployed (e.g., waiting for a defined delay).
See PR #1437
Enhanced Conflict Resolution and Metadata Handling
The new Tier field has been added to PolicyRef and KustomizationRef objects. This introduces a secondary, intra-profile priority system designed to resolve resource conflicts.
When multiple ClusterProfile or Profile objects targeting the same cluster attempt to deploy the identical Kubernetes resource, the ClusterProfile with the lowest numerical Tier value wins (lower number indicates higher priority).
See PR #1439
Resolved an issue that caused unnecessary reconciliation loops for add-on profiles, leading to excessive resource consumption and potential rate-limiting. The controller now only reconciles when actual changes occur.
Fixed a critical bug that caused the controller to crash when attempting to process a ConfigurationGroup resource whose status field was empty. This improves the overall robustness and uptime of the controller.
When the templating process for add-ons fails (e.g., missing values), the controller now accurately reports the specific error in the ClusterSummary Status. This significantly streamlines troubleshooting for failed deployments.
Fixed an issue where the ContinuousWithDriftDetection synchronization mode did not work correctly when used with KustomizationRefresources.
The ClusterSummary status now explicitly reports when a deployment is halted because either the target Cluster or the associated Profile/ClusterProfile is paused. This provides clearer feedback on the current operational state.
Refined the ValuesFrom feature to ensure that template instantiation is only attempted for ConfigMap or Secret data that includes the explicit projectsveltos.io/template annotation. This prevents unintended template parsing and potential errors.
General improvements to error reporting within the ClusterSummary resource, providing more detailed and actionable information about deployment failures. This enhancement makes it easier to diagnose and resolve add-on deployment issues across your fleet.
To remove the previous limit of 63 characters on resource names, Sveltos now stores referenced resource metadata as annotations instead of labels. This resolves limitations related to label value length.
Added a check to the Cluster Predicate to detect the ClusterAPI Cluster's ControlPlaneInitialized condition is set to True. This prevents unnecessary delays deploying add-ons.
10 . Reduced Redundant EventReport Updates: PR 570
The system now sorts and maintains matching resources within the EventReport to prevent redundant updates. Previously, an EventSource watcher on a resource could trigger reprocessing and an EventReport update even if the list of matching resources hadn't actually changed (especially when collectResources was false), leading to unnecessary work on the control cluster. Skipping the update when the sorted list is identical prevents this issue.
Nothing published for this version
A new Custom Resource Definition (ClusterPromotion) is introduced to enable declarative, multi-stage, progressive rollout of a single configuration (p
A new Custom Resource Definition (ClusterPromotion) is introduced to enable declarative, multi-stage, progressive rollout of a single configuration (profileSpec) across multiple sets of clusters. This is ideal for promoting configurations from environments like staging to production in a controlled, sequential manner.
It manages an ordered series of stages, each with its own cluster selector.
It supports automated promotion to the next stage after the previous stage's configuration is successfully deployed (e.g., waiting for a defined delay).
See PR #1437
Enhanced Conflict Resolution and Metadata Handling
The new Tier field has been added to PolicyRef and KustomizationRef objects. This introduces a secondary, intra-profile priority system designed to resolve resource conflicts.
When multiple ClusterProfile or Profile objects targeting the same cluster attempt to deploy the identical Kubernetes resource, the ClusterProfile with the lowest numerical Tier value wins (lower number indicates higher priority).
See PR #1439
Resolved an issue that caused unnecessary reconciliation loops for add-on profiles, leading to excessive resource consumption and potential rate-limiting. The controller now only reconciles when actual changes occur.
Fixed a critical bug that caused the controller to crash when attempting to process a ConfigurationGroup resource whose status field was empty. This improves the overall robustness and uptime of the controller.
When the templating process for add-ons fails (e.g., missing values), the controller now accurately reports the specific error in the ClusterSummary Status. This significantly streamlines troubleshooting for failed deployments.
Fixed an issue where the ContinuousWithDriftDetection synchronization mode did not work correctly when used with KustomizationRefresources.
The ClusterSummary status now explicitly reports when a deployment is halted because either the target Cluster or the associated Profile/ClusterProfile is paused. This provides clearer feedback on the current operational state.
Refined the ValuesFrom feature to ensure that template instantiation is only attempted for ConfigMap or Secret data that includes the explicit projectsveltos.io/template annotation. This prevents unintended template parsing and potential errors.
General improvements to error reporting within the ClusterSummary resource, providing more detailed and actionable information about deployment failures. This enhancement makes it easier to diagnose and resolve add-on deployment issues across your fleet.
To remove the previous limit of 63 characters on resource names, Sveltos now stores referenced resource metadata as annotations instead of labels. This resolves limitations related to label value length.
Added a check to the Cluster Predicate to detect the ClusterAPI Cluster's ControlPlaneInitialized condition is set to True. This prevents unnecessary delays deploying add-ons.
We fixed a bug on the MCP Server and improved our dashboard's debug reporting.
We fixed a bug on the MCP Server and improved our dashboard's debug reporting.
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
One column per quarter.