io.druid.indexing.overlord.autoscaling

Interface ResourceManagementStrategy<T extends TaskRunner>

All Known Implementing Classes:

AbstractWorkerResourceManagementStrategy, NoopResourceManagementStrategy, PendingTaskBasedWorkerResourceManagementStrategy, SimpleWorkerResourceManagementStrategy

public interface ResourceManagementStrategy<T extends TaskRunner>

The ResourceManagementStrategy decides if worker nodes should be provisioned or determined based on the available tasks in the system and the state of the workers in the system. In general, the resource management is tied to the runner.

Method Summary

Modifier and Type	Method and Description
ScalingStats	getStats() Get any interesting stats related to scaling
void	startManagement(T runner) Equivalent to start() but requires a specific runner instance which holds state of interest.
void	stopManagement() Equivalent to stop() Should be called from TaskRunner's lifecycle

Method Detail

startManagement

void startManagement(T runner)

Equivalent to start() but requires a specific runner instance which holds state of interest. This method is intended to be called from the TaskRunner's lifecycle

Parameters:

runner - The TaskRunner state holder this strategy should use during execution

stopManagement

void stopManagement()

Equivalent to stop() Should be called from TaskRunner's lifecycle

getStats

ScalingStats getStats()

Get any interesting stats related to scaling

Returns:

The ScalingStats or `null` if nothing of interest