Cluster workloads¶

Mounted at client.clusters.workloads.

barbara.resources.workloads.ClusterWorkloadsResource ¶

ClusterWorkloadsResource(client: BarbaraClient)

create_marketplace_workload ¶

create_marketplace_workload(
    cluster_id: str,
    *,
    app_version_id: str,
    application_id: str,
    name: str,
    force_pull: bool = False,
    enable_logs: bool = True,
    gpu: Optional[bool] = None,
    config: Optional[Dict[str, Any]] = None,
    config_id: Optional[str] = None,
    compose_config: Optional[List[Dict[str, Any]]] = None,
    app_secrets: Optional[List[Dict[str, Any]]] = None,
    deployment: Optional[List[Dict[str, Any]]] = None,
    placement_constraints: Optional[
        List[Dict[str, Any]]
    ] = None,
) -> None

deployment ({"name", "mode", "replicas"} per service) and placement_constraints ({"name", "constraints"} per service) are cluster-only — see _build_stack_services for the full semantics of each.

update_marketplace_workload ¶

update_marketplace_workload(
    cluster_id: str,
    workload_id: str,
    *,
    app_version_id: str,
    application_id: str,
    force_pull: bool = False,
    enable_logs: bool = True,
    gpu: Optional[bool] = None,
    config: Optional[Dict[str, Any]] = None,
    config_id: Optional[str] = None,
    name: Optional[str] = None,
    compose_config: Optional[List[Dict[str, Any]]] = None,
    app_secrets: Optional[List[Dict[str, Any]]] = None,
    deployment: Optional[List[Dict[str, Any]]] = None,
    placement_constraints: Optional[
        List[Dict[str, Any]]
    ] = None,
) -> None

Cluster-level workloads have no get() to read back their current state — unlike the node-level equivalent, giving only some of compose_config/app_secrets/deployment/ placement_constraints here REPLACES the whole services array and clears whichever axes you didn't pass. Track and pass the full desired state yourself if that matters.

update_marketplace_workload_compose_config ¶

update_marketplace_workload_compose_config(
    cluster_id: str,
    workload_id: str,
    compose_config: List[Dict[str, Any]],
) -> None

No get() at cluster scope (see update_marketplace_workload) — this REPLACES the entire services array, clearing App Secrets, Deployment & Replicas, and Placement Constraints if not included here.

update_marketplace_workload_app_secrets ¶

update_marketplace_workload_app_secrets(
    cluster_id: str,
    workload_id: str,
    app_secrets: List[Dict[str, Any]],
) -> None

No get() at cluster scope (see update_marketplace_workload) — this REPLACES the entire services array, clearing Compose Config, Deployment & Replicas, and Placement Constraints if not included here.

get_urls ¶

get_urls(
    cluster_id: str, workload_id: str
) -> List[Dict[str, Any]]

Reads back the URLs Panel shows for a deployed Marketplace app. There's no per-workload get() at cluster scope, but client.clusters.get(cluster_id) embeds every stack's full current state under raw["stacks"], including this at stacks[].app.current.marketAppVersionId.uris. See barbara.utils.decode_uris for the shape.

get_running_services ¶

get_running_services(
    cluster_id: str, workload_id: str
) -> List[Dict[str, Any]]

Reads back the live Docker Swarm service state backing this stack — stacks[].clusterInfo.services[], one entry per Swarm service with its own tasks[] (per-node running/shutdown/failed/ orphaned state). Returned as-is (not base64-decoded — this is Docker's own JSON, unlike Compose Config/App Secrets), unlike the node-level get_running_services() which is a lightly-normalized subset of Docker container info.

create_model_workload ¶

create_model_workload(
    cluster_id: str,
    *,
    app_version_id: str,
    application_id: str,
    name: str,
    force_pull: bool = False,
    enable_logs: bool = True,
    gpu: Optional[bool] = None,
    compose_config: Optional[List[Dict[str, Any]]] = None,
    deployment: Optional[List[Dict[str, Any]]] = None,
    placement_constraints: Optional[
        List[Dict[str, Any]]
    ] = None,
) -> None

Model workloads have Compose Config, Deployment & Replicas, and Placement Constraints, but neither App Config nor App Secrets — see NodeWorkloadsResource.create_model_workload. compose_config must exactly match the service template declared by the selected model application version.

update_model_workload_compose_config ¶

update_model_workload_compose_config(
    cluster_id: str,
    workload_id: str,
    compose_config: List[Dict[str, Any]],
) -> None

No get() at cluster scope — this REPLACES the entire services array, clearing Deployment & Replicas and Placement Constraints if not included here.

barbara.resources.workloads.AsyncClusterWorkloadsResource ¶

AsyncClusterWorkloadsResource(client: AsyncBarbaraClient)

update_marketplace_workload `async` ¶

update_marketplace_workload(
    cluster_id: str,
    workload_id: str,
    *,
    app_version_id: str,
    application_id: str,
    force_pull: bool = False,
    enable_logs: bool = True,
    gpu: Optional[bool] = None,
    config: Optional[Dict[str, Any]] = None,
    config_id: Optional[str] = None,
    name: Optional[str] = None,
    compose_config: Optional[List[Dict[str, Any]]] = None,
    app_secrets: Optional[List[Dict[str, Any]]] = None,
    deployment: Optional[List[Dict[str, Any]]] = None,
    placement_constraints: Optional[
        List[Dict[str, Any]]
    ] = None,
) -> None

Cluster-level workloads have no get() to read back their current state — unlike the node-level equivalent, giving only some of compose_config/app_secrets/deployment/ placement_constraints here REPLACES the whole services array and clears whichever axes you didn't pass.

update_marketplace_workload_compose_config `async` ¶

update_marketplace_workload_compose_config(
    cluster_id: str,
    workload_id: str,
    compose_config: List[Dict[str, Any]],
) -> None