Workloads

A Workload defines a scoped view of your HPC job data: which clusters to query and what filters to apply when selecting jobs. Features such as Healthcheck use workloads to restrict the data they operate on; other features (e.g., reporting) will use them in the same way.

Workloads are managed from Management → Workloads.

Workload list

The Workloads list page shows all existing workloads with their name, linked clusters, and the number of features currently referencing each workload.

workload_list

From this page you can create, edit, or delete workloads.

Note

A workload that is referenced by any OKA feature cannot be deleted. Remove or reassign all references to it first.

Creating or editing a workload

workload_form

The workload form has three sections: identity, cluster selection, and filters.

Identity

  • Name (required) — A unique name for the workload.

  • Description — An optional free-text description of what this workload represents.

Cluster selection

  • Clusters (required) — One or more clusters whose data this workload will query. Select from the list of all clusters configured in OKA.

Filters

Filters restrict the job data that becomes available when this workload is evaluated. Two independent filter sets are available, one per data source:

  • Accounting — filters applied to accounting (job history) data.

  • Monitoring — filters applied to monitoring (node occupancy) data.

workload_filters

Each filter set has three independent components that can be combined:

1. Date range

Choose one of three modes:

  • Period — a named rolling window (e.g., Last 30 days, Month to date, Previous month). The window is resolved to absolute dates each time the workload is evaluated.

    Available periods:

    Value

    Description

    last_30_days

    Rolling last 30 days

    last_6_months

    Rolling last 6 months

    last_year

    Rolling last 12 months

    month_to_date

    From the 1st of the current month to now

    quarter_to_date

    From the start of the current quarter to now

    semester_to_date

    From the start of the current half-year to now

    year_to_date

    From January 1st of the current year to now

    previous_month

    The full previous calendar month

    previous_quarter

    The full previous calendar quarter

    previous_semester

    The full previous half-year

    previous_year

    The full previous calendar year

    all

    No date filter (all available data)

  • Custom period (days) — a rolling window of N days back from the moment of evaluation.

  • Datetime range — an explicit start and end date, selected via a calendar picker. Use this for historical or fixed-window analysis.

2. Advanced filter

An advanced filter component for applying field-level conditions (e.g., restrict to a specific partition, account, or job state). Conditions within a group or between groups can be combined with AND or OR conditions.

If the selected clusters have no data sources configured, a warning is shown and the advanced filter is unavailable until a data source is set up.

3. Recurring filter

Restricts evaluation to jobs submitted during specific times of day or days of the week.

  • Start time / End time — the time window within a day (default: 00:00–23:59).

  • Days — checkboxes for Monday through Sunday (all selected by default).

Note

All three filter components within a source tab (date range, advanced, recurring) are applied simultaneously as an AND condition. The accounting and monitoring filter sets are independent of each other.