Release Notes

Breaking changes

• Removed `DeploymentHealth` class from `ui/health-checker`
  The `DeploymentHealth` class and its `to_dict` method have been removed from the `ui/health-checker` module. Any tooling or automation that consumed `DeploymentHealth` objects or relied on their serialized output will break.
  Replace usages of `DeploymentHealth` with the new `LayerHealth` class, which provides equivalent health status representation with additional layered context. Update any code that calls `DeploymentHealth.to_dict()` to use `LayerHealth.to_dict()` instead.
• Removed DRBD operator agent internal modules (`device_discovery`, `drbd_manager`, `health_monitor`)
  The `DeviceDiscovery`, `DRBDManager`, and `HealthMonitor` classes — along with all their associated methods — have been removed from `drbd-operator/src/agent`. This includes over 475 additional removals across the agent layer. Any integrations or extensions that imported or subclassed these components will no longer function.
  If you have custom tooling that directly interfaces with `DeviceDiscovery`, `DRBDManager`, or `HealthMonitor`, you will need to migrate to the new health-checker framework (`HealthChecker`, `LinstorClient`, `ClusterClients`). Review the updated `ui/health-checker` module APIs and replace direct agent interactions with the provided client abstractions. Consult the updated operator documentation for the new architecture.

New features

• tsr-gather: Must-Gather Diagnostics Collector
  Introduced `tsr-gather`, a must-gather compatible diagnostics collection tool with quorum reach-through support. Platform engineers and SREs can now collect comprehensive cluster diagnostics — including site recovery state, quorum status, and storage health — in a single operation, simplifying troubleshooting of disaster recovery scenarios.
• Layered Health-Check Framework (`HealthChecker`)
  Added a new `HealthChecker` class with `run_all_checks()`, `run_layer1_checks()`, and additional per-layer check methods. Health results are now modeled as `HealthCheckResult` and `LayerHealth` objects, each serializable via `to_dict()`, enabling structured reporting across multiple layers of the recovery stack.
• LINSTOR Storage Client (`LinstorClient`)
  Added `LinstorClient` to the `ui/health-checker` module, providing programmatic access to LINSTOR storage operations including querying resource groups, resource definitions, storage pools, nodes, and resource volumes, as well as modifying resource groups. This enables health checks to incorporate deep storage-layer insights.
• Cluster Client Initialization (`ClusterClients`)
  Added `ClusterClients` with an `initialize()` method to centrally manage and configure connections to cluster-level clients (including LINSTOR) used during health checking and diagnostics collection.
• Structured Remediation Suggestions (`Remediation`)
  Added `Remediation` and `RemediationCommand` classes to the `ui/health-checker` module. Health check results can now include actionable remediation steps, each serializable via `to_dict()`, making it easier to surface recovery guidance directly from diagnostics output.

API surface