# Scheduling The operator provides three custom resources for scheduling tasks in Plakar Control Plane: - `ScheduleBackup` - `ScheduleCheck` - `ScheduleSync` These resources are the declarative equivalents of the scheduled tasks that can be configured through the [Scheduling](../../../scheduling/) section of Plakar Control Plane. All three resources use the same `periodicity` field, which specifies the execution interval in seconds. Rather than referencing Plakar Control Plane resource UUIDs, they reference other Kubernetes custom resources by name, and optionally by namespace. They also share an optional `historyLimit` field, covered in [Job history](#job-history) below. ## ScheduleBackup A `ScheduleBackup` resource creates a recurring backup task for a [Source](../sources) and stores the backups in a [Store](../stores). ```yaml apiVersion: task.plakar.io/v1alpha1 kind: ScheduleBackup metadata: name: nightly-database-backup spec: periodicity: 86400 source: name: production-database store: name: production-s3 ``` ## ScheduleCheck A `ScheduleCheck` resource periodically verifies the integrity of a [Store](../stores). ```yaml apiVersion: task.plakar.io/v1alpha1 kind: ScheduleCheck metadata: name: weekly-store-check spec: periodicity: 604800 store: name: production-s3 ``` ## ScheduleSync A `ScheduleSync` resource periodically copies backup data from one [Store](../stores) to another. ```yaml apiVersion: task.plakar.io/v1alpha1 kind: ScheduleSync metadata: name: replicate-to-cold-storage spec: periodicity: 86400 sourceStore: name: production-s3 destinationStore: name: cold-storage-glacier ``` ## Running tasks on a remote edge All three resources also accept an optional `edgeTags` field that restricts the task to [edges](../../../infrastructure/edges) reporting the listed tags: ```yaml apiVersion: task.plakar.io/v1alpha1 kind: ScheduleBackup metadata: name: nightly-database-backup spec: periodicity: 86400 source: name: production-database store: name: production-s3 edgeTags: - gpu - eu-west ``` When `edgeTags` is set, the task runs on an online edge that reports every listed tag. It fails if no online edge matches, and never runs on the Control Plane. When `edgeTags` is omitted, the task runs on any online edge, and only runs on the Control Plane itself when no edge is online. See [Selecting an edge](../../../infrastructure/edges#selecting-an-edge). After a scheduling resource is created, the corresponding scheduled task is created in Plakar Control Plane. Its UUID is exposed through `status.id`, while `status.conditions` reports whether the task was created successfully. ## Job history Each execution of a `ScheduleBackup`, `ScheduleCheck`, or `ScheduleSync` is mirrored in the cluster as a `TaskRun` resource, in the same namespace as the schedule that produced it. A `TaskRun` reports the execution's phase (`queued`, `running`, `canceled`, `succeeded`, or `failed`) and its scheduled, started, and stopped timestamps: ```sh $ kubectl get taskruns ``` `TaskRun` resources are read-only records owned by the schedule that created them, so deleting a `ScheduleBackup`, `ScheduleCheck`, or `ScheduleSync` resource also deletes its `TaskRun` history. Each one carries `task.plakar.io/name` and `task.plakar.io/kind` labels identifying that parent, so you can list only the runs for a given schedule: ```sh $ kubectl get taskruns -l task.plakar.io/name=backup ``` ```txt NAME TASK KIND PHASE SCHEDULEDAT AGE backup-0d2a6cd0 backup schedulebackup running 16s 16s backup-609ae284 backup schedulebackup succeeded 106s 79s backup-917e49c9 backup schedulebackup succeeded 46s 17s backup-a241c8ff backup schedulebackup succeeded 2m16s 110s backup-f1788ce4 backup schedulebackup succeeded 76s 48s ``` `kubectl describe` on a single `TaskRun` shows its `jobID`, the `ScheduleBackup`, `ScheduleCheck`, or `ScheduleSync` resource that owns it, and the timestamps reported by Plakar Control Plane for that execution: ```sh $ kubectl describe taskrun backup-0d2a6cd0 ``` ```txt Name: backup-0d2a6cd0 Namespace: default Labels: task.plakar.io/kind=schedulebackup task.plakar.io/name=backup Annotations: API Version: task.plakar.io/v1alpha1 Kind: TaskRun Metadata: Creation Timestamp: 2026-08-06T09:00:05Z Generation: 1 Owner References: API Version: task.plakar.io/v1alpha1 Block Owner Deletion: true Controller: true Kind: ScheduleBackup Name: backup UID: ea5004e0-18af-42ab-8d4f-543e3afe8ad0 Resource Version: 112969151773 UID: 5df1b84f-bb6a-4067-8630-ea757190fded Spec: Job ID: 0d2a6cd0-66a2-468c-9040-48f8995c9f7e Task Ref: Kind: schedulebackup Name: backup Status: Phase: running Scheduled At: 2026-08-06T09:00:05Z Started At: 2026-08-06T09:00:05Z Events: ``` `TaskRun` resources mirror the job history reported by Plakar Control Plane. They are not a permanent record. When Plakar Control Plane stops reporting a job, the operator prunes its `TaskRun`. The `historyLimit` field caps how many `TaskRun` resources are kept for a given schedule. It defaults to 5 and can be set up to 20: ```yaml apiVersion: task.plakar.io/v1alpha1 kind: ScheduleBackup metadata: name: nightly-database-backup spec: periodicity: 86400 historyLimit: 10 source: name: production-database store: name: production-s3 ``` The operator authenticates as an Application user, and creates a scheduler for that user in Plakar Control Plane if one doesn't already exist. Every `ScheduleBackup`, `ScheduleCheck`, and `ScheduleSync` resource it manages then appears under that user's entry in the **User schedulers** view described in [User Schedulers](../../../scheduling/user-schedulers). These resources only cover recurring tasks. For one-off restores, see the [Restore Resource](../restore). For any other task run once, use the [manual scheduler](../../../scheduling/manual-scheduler).