Struct ClusterController

pub struct ClusterController {
    env: MetaSrvEnv,
    max_heartbeat_interval: Duration,
    inner: RwLock<ClusterControllerInner>,
    started_at: u64,
}

Fields

env: MetaSrvEnv

max_heartbeat_interval: Duration

inner: RwLock<ClusterControllerInner>

started_at: u64

Used as timestamp when meta node starts in sec.

Implementations

impl ClusterController

pub async fn new( env: MetaSrvEnv, max_heartbeat_interval: Duration, ) -> MetaResult<Self>

pub async fn get_inner_read_guard( &self, ) -> RwLockReadGuard<'_, ClusterControllerInner>

Used in NotificationService::subscribe. Need to pay attention to the order of acquiring locks to prevent deadlock problems.

pub async fn count_worker_by_type(&self) -> MetaResult<HashMap<WorkerType, i64>>

pub async fn add_worker( &self, type: PbWorkerType, host_address: HostAddress, property: AddNodeProperty, resource: PbResource, ) -> MetaResult<WorkerId>

A worker node will immediately register itself to meta when it bootstraps. The meta will assign it with a unique ID and set its state as Starting. When the worker node is fully ready to serve, it will request meta again (via activate_worker_node) to set its state to Running.

pub async fn activate_worker(&self, worker_id: WorkerId) -> MetaResult<()>

pub async fn delete_worker( &self, host_address: HostAddress, ) -> MetaResult<WorkerNode>

pub async fn update_schedulability( &self, worker_ids: Vec<WorkerId>, schedulability: Schedulability, ) -> MetaResult<()>

pub async fn heartbeat(&self, worker_id: WorkerId) -> MetaResult<()>

Invoked when it receives a heartbeat from a worker node.

pub fn start_heartbeat_checker( cluster_controller: ClusterControllerRef, check_interval: Duration, ) -> (JoinHandle<()>, Sender<()>)

pub async fn list_workers( &self, worker_type: Option<WorkerType>, worker_status: Option<WorkerStatus>, ) -> MetaResult<Vec<PbWorkerNode>>

Get live nodes with the specified type and state.

Arguments

• worker_type WorkerType of the nodes
• worker_state Filter by this state if it is not None.

pub(crate) async fn subscribe_active_streaming_compute_nodes( &self, ) -> MetaResult<(Vec<WorkerNode>, UnboundedReceiver<LocalNotification>)>

pub async fn list_active_streaming_workers( &self, ) -> MetaResult<Vec<PbWorkerNode>>

A convenient method to get all running compute nodes that may have running actors on them i.e. CNs which are running

pub async fn list_active_worker_slots(&self) -> MetaResult<Vec<WorkerSlotId>>

pub async fn list_active_serving_workers(&self) -> MetaResult<Vec<PbWorkerNode>>

Get the cluster info used for scheduling a streaming job, containing all nodes that are running and schedulable

pub async fn get_streaming_cluster_info( &self, ) -> MetaResult<StreamingClusterInfo>

Get the cluster info used for scheduling a streaming job.

pub async fn get_worker_by_id( &self, worker_id: WorkerId, ) -> MetaResult<Option<PbWorkerNode>>

pub async fn get_worker_info_by_id( &self, worker_id: WorkerId, ) -> Option<WorkerExtraInfo>

pub fn cluster_id(&self) -> &ClusterId