Plakar Operator

New

• Added Restore CRD type for orchestrating restore operations.
• Added TaskRun CRD to report the status of individual task executions on Kubernetes.
• Added HistoryLimit field to ScheduleBackup, ScheduleSync, ScheduleCheck, and Restore CRDs to control how many TaskRun records are retained (default: 5).
• Added EdgeTags to ScheduleBackup, ScheduleSync, ScheduleCheck, and Restore.
• Added observedGeneration to connector specs.
• Connectors are tested for reachability on create and update.

Improvements

• Sync is skipped when the connector generation is unchanged, avoiding redundant reconciliation cycles.
• Plakar.spec.inventoryUUID is now optional; the operator creates one automatically when it is unset.
• The operator uses the user-configured scheduler instead of hardcoding "custom".

Breaking changes

• Version field removed from the Integration CRD spec.
• Only a single Plakar resource is accepted per namespace for now; extra resources are rejected.

New

• Added Destination CRD for destination connectors.
• Added FieldsFrom and FieldValue fields to Source and Store specs.
• Introduced ScheduleSync, ScheduleCheck, and Restore CRDs for task scheduling.
• Added multi-group support for CRD organisation.
• Scaffolded a distributable Helm chart.
• Login to the Plakar control plane now uses API keys.
• Added API reference documentation, generated and published in CI.
• Sample manifests added for every CRD.

Improvements

• Connector reconciliation logic refactored into a shared generic function.
• Connector spec updates are now propagated to the control plane via UpdateConnector.
• ScheduleBackup updates the task in the control plane when the spec changes.
• Replaced Has + Update calls with a single SyncConnector to avoid a redundant GET.
• plakman: actor is now set for scheduler interactions.