org.apache.flink.runtime.jobmaster.slotpool

Class SlotPoolImpl

java.lang.Object

org.apache.flink.runtime.jobmaster.slotpool.SlotPoolImpl

All Implemented Interfaces:

AutoCloseable, AllocatedSlotActions, SlotPool, SlotPoolService

public class SlotPoolImpl
extends Object
implements SlotPool, SlotPoolService

The slot pool serves slot request issued by ExecutionGraph. It will attempt to acquire new slots from the ResourceManager when it cannot serve a slot request. If no ResourceManager is currently available, or it gets a decline from the ResourceManager, or a request times out, it fails the slot request. The slot pool also holds all the slots that were offered to it and accepted, and can thus provides registered free slots even if the ResourceManager is down. The slots will only be released when they are useless, e.g. when the job is fully running but we still have some free slots.

All the allocation or the slot offering will be identified by self generated AllocationID, we will use it to eliminate ambiguities.

TODO : Make pending requests location preference aware TODO : Make pass location preferences to ResourceManager when sending a slot request

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
protected static class	SlotPoolImpl.AvailableSlots Organize all available slots from different points of view.
protected static class	SlotPoolImpl.PendingRequest A pending request for a slot.

Field Summary

Fields

Modifier and Type	Field and Description
protected boolean	batchSlotRequestTimeoutCheckEnabled
protected org.slf4j.Logger	log

Constructor Summary

Constructors

Constructor and Description
SlotPoolImpl(org.apache.flink.api.common.JobID jobId, org.apache.flink.util.clock.Clock clock, org.apache.flink.api.common.time.Time rpcTimeout, org.apache.flink.api.common.time.Time idleSlotTimeout, org.apache.flink.api.common.time.Time batchSlotTimeout)

Method Summary

Modifier and Type	Method and Description
Optional<PhysicalSlot>	allocateAvailableSlot(SlotRequestId slotRequestId, AllocationID allocationID, ResourceProfile requirementProfile) Allocates the available slot with the given allocation id under the given request id for the given requirement profile.
protected void	checkBatchSlotTimeout()
protected void	checkIdleSlot() Check the available slots, release the slot that is idle for a long time.
void	close() Close the slot pool service.
void	connectToResourceManager(ResourceManagerGateway resourceManagerGateway) Connects the SlotPool to the given ResourceManager.
AllocatedSlotReport	createAllocatedSlotReport(ResourceID taskManagerId) Create report about the allocated slots belonging to the specified task manager.
void	disableBatchSlotRequestTimeoutCheck() Disables batch slot request timeout check.
void	disconnectResourceManager() Disconnects the slot pool from its current Resource Manager.
Optional<ResourceID>	failAllocation(AllocationID allocationID, Exception cause) Fail the specified allocation and release the corresponding slot if we have one.
Optional<ResourceID>	failAllocation(ResourceID resourceId, AllocationID allocationID, Exception cause) Fails the allocation with the given allocationId.
Collection<SlotInfo>	getAllocatedSlotsInformation() Returns a list of SlotInfo objects about all slots that are currently allocated in the slot pool.
Collection<SlotInfoWithUtilization>	getAvailableSlotsInformation() Returns a list of SlotInfoWithUtilization objects about all slots that are currently available in the slot pool.
DualKeyLinkedMap<SlotRequestId,AllocationID,SlotPoolImpl.PendingRequest>	getPendingRequests()
Collection<SlotOffer>	offerSlots(TaskManagerLocation taskManagerLocation, TaskManagerGateway taskManagerGateway, Collection<SlotOffer> offers) Offers multiple slots to the SlotPool.
boolean	registerTaskManager(ResourceID resourceID) Register TaskManager to this pool, only those slots come from registered TaskManager will be considered valid.
void	releaseSlot(SlotRequestId slotRequestId, Throwable cause) Releases the slot with the given SlotRequestId.
boolean	releaseTaskManager(ResourceID resourceId, Exception cause) Unregister TaskManager from this pool, all the related slots will be released and tasks be canceled.
CompletableFuture<PhysicalSlot>	requestNewAllocatedBatchSlot(SlotRequestId slotRequestId, ResourceProfile resourceProfile) Requests the allocation of a new batch slot from the resource manager.
CompletableFuture<PhysicalSlot>	requestNewAllocatedSlot(SlotRequestId slotRequestId, ResourceProfile resourceProfile, org.apache.flink.api.common.time.Time timeout) Request the allocation of a new slot from the resource manager.
protected void	runAsync(Runnable runnable) Execute the runnable in the main thread of the underlying RPC endpoint.
protected void	scheduleRunAsync(Runnable runnable, long delay, TimeUnit unit) Execute the runnable in the main thread of the underlying RPC endpoint, with a delay of the given number of milliseconds.
protected void	scheduleRunAsync(Runnable runnable, org.apache.flink.api.common.time.Time delay) Execute the runnable in the main thread of the underlying RPC endpoint, with a delay of the given number of milliseconds.
void	start(JobMasterId jobMasterId, String newJobManagerAddress, ComponentMainThreadExecutor componentMainThreadExecutor) Start the slot pool to accept RPC calls.
protected void	timeoutPendingSlotRequest(SlotRequestId slotRequestId)

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Methods inherited from interface org.apache.flink.runtime.jobmaster.slotpool.SlotPoolService

castInto, notifyNotEnoughResourcesAvailable

Field Detail

log

protected final org.slf4j.Logger log

batchSlotRequestTimeoutCheckEnabled

protected boolean batchSlotRequestTimeoutCheckEnabled

Constructor Detail

SlotPoolImpl

public SlotPoolImpl(org.apache.flink.api.common.JobID jobId,
                    org.apache.flink.util.clock.Clock clock,
                    org.apache.flink.api.common.time.Time rpcTimeout,
                    org.apache.flink.api.common.time.Time idleSlotTimeout,
                    org.apache.flink.api.common.time.Time batchSlotTimeout)

Method Detail

getAllocatedSlotsInformation

public Collection<SlotInfo> getAllocatedSlotsInformation()

Description copied from interface: SlotPool

Returns a list of SlotInfo objects about all slots that are currently allocated in the slot pool.

Specified by:

getAllocatedSlotsInformation in interface SlotPool

Returns:

a list of SlotInfo objects about all slots that are currently allocated in the slot pool.

getPendingRequests

@VisibleForTesting
public DualKeyLinkedMap<SlotRequestId,AllocationID,SlotPoolImpl.PendingRequest> getPendingRequests()

start

public void start(@Nonnull
                  JobMasterId jobMasterId,
                  @Nonnull
                  String newJobManagerAddress,
                  @Nonnull
                  ComponentMainThreadExecutor componentMainThreadExecutor)
           throws Exception

Start the slot pool to accept RPC calls.

Specified by:

start in interface SlotPool

Specified by:

start in interface SlotPoolService

Parameters:

jobMasterId - The necessary leader id for running the job.

newJobManagerAddress - for the slot requests which are sent to the resource manager

componentMainThreadExecutor - The main thread executor for the job master's main thread.

Throws:

Exception - if the the service cannot be started

close

public void close()

Description copied from interface: SlotPoolService

Close the slot pool service.

Specified by:

close in interface AutoCloseable

Specified by:

close in interface SlotPool

Specified by:

close in interface SlotPoolService

connectToResourceManager

public void connectToResourceManager(@Nonnull
                                     ResourceManagerGateway resourceManagerGateway)

Description copied from interface: SlotPool

Connects the SlotPool to the given ResourceManager. After this method is called, the SlotPool will be able to request resources from the given ResourceManager.

Specified by:

connectToResourceManager in interface SlotPool

Specified by:

connectToResourceManager in interface SlotPoolService

Parameters:

resourceManagerGateway - The RPC gateway for the resource manager.

disconnectResourceManager

public void disconnectResourceManager()

Description copied from interface: SlotPool

Disconnects the slot pool from its current Resource Manager. After this call, the pool will not be able to request further slots from the Resource Manager, and all currently pending requests to the resource manager will be canceled.

The slot pool will still be able to serve slots from its internal pool.

Specified by:

disconnectResourceManager in interface SlotPool

Specified by:

disconnectResourceManager in interface SlotPoolService

releaseSlot

public void releaseSlot(@Nonnull
                        SlotRequestId slotRequestId,
                        @Nullable
                        Throwable cause)

Description copied from interface: AllocatedSlotActions

Releases the slot with the given SlotRequestId. Additionally, one can provide a cause for the slot release.

Specified by:

releaseSlot in interface AllocatedSlotActions

Parameters:

slotRequestId - identifying the slot to release

cause - of the slot release, null if none

allocateAvailableSlot

public Optional<PhysicalSlot> allocateAvailableSlot(@Nonnull
                                                    SlotRequestId slotRequestId,
                                                    @Nonnull
                                                    AllocationID allocationID,
                                                    @Nonnull
                                                    ResourceProfile requirementProfile)

Description copied from interface: SlotPool

Allocates the available slot with the given allocation id under the given request id for the given requirement profile. The slot must be able to fulfill the requirement profile, otherwise an IllegalStateException will be thrown.

Specified by:

allocateAvailableSlot in interface SlotPool

Parameters:

slotRequestId - identifying the requested slot

allocationID - the allocation id of the requested available slot

requirementProfile - resource profile of the requirement for which to allocate the slot

Returns:

the previously available slot with the given allocation id, if a slot with this allocation id exists

requestNewAllocatedSlot

@Nonnull
public CompletableFuture<PhysicalSlot> requestNewAllocatedSlot(@Nonnull
                                                                        SlotRequestId slotRequestId,
                                                                        @Nonnull
                                                                        ResourceProfile resourceProfile,
                                                                        @Nullable
                                                                        org.apache.flink.api.common.time.Time timeout)

Description copied from interface: SlotPool

Request the allocation of a new slot from the resource manager. This method will not return a slot from the already available slots from the pool, but instead will add a new slot to that pool that is immediately allocated and returned.

Specified by:

requestNewAllocatedSlot in interface SlotPool

Parameters:

slotRequestId - identifying the requested slot

resourceProfile - resource profile that specifies the resource requirements for the requested slot

timeout - timeout for the allocation procedure

Returns:

a newly allocated slot that was previously not available.

requestNewAllocatedBatchSlot

@Nonnull
public CompletableFuture<PhysicalSlot> requestNewAllocatedBatchSlot(@Nonnull
                                                                             SlotRequestId slotRequestId,
                                                                             @Nonnull
                                                                             ResourceProfile resourceProfile)

Description copied from interface: SlotPool

Requests the allocation of a new batch slot from the resource manager. Unlike the normal slot, a batch slot will only time out if the slot pool does not contain a suitable slot. Moreover, it won't react to failure signals from the resource manager.

Specified by:

requestNewAllocatedBatchSlot in interface SlotPool

Parameters:

slotRequestId - identifying the requested slot

resourceProfile - resource profile that specifies the resource requirements for the requested batch slot

Returns:

a future which is completed with newly allocated batch slot

disableBatchSlotRequestTimeoutCheck

public void disableBatchSlotRequestTimeoutCheck()

Description copied from interface: SlotPool

Disables batch slot request timeout check. Invoked when someone else wants to take over the timeout check responsibility.

Specified by:

disableBatchSlotRequestTimeoutCheck in interface SlotPool

getAvailableSlotsInformation

@Nonnull
public Collection<SlotInfoWithUtilization> getAvailableSlotsInformation()

Description copied from interface: SlotPool

Returns a list of SlotInfoWithUtilization objects about all slots that are currently available in the slot pool.

Specified by:

getAvailableSlotsInformation in interface SlotPool

Returns:

a list of SlotInfoWithUtilization objects about all slots that are currently available in the slot pool.

offerSlots

public Collection<SlotOffer> offerSlots(TaskManagerLocation taskManagerLocation,
                                        TaskManagerGateway taskManagerGateway,
                                        Collection<SlotOffer> offers)

Description copied from interface: SlotPool

Offers multiple slots to the SlotPool. The slot offerings can be individually accepted or rejected by returning the collection of accepted slot offers.

Specified by:

offerSlots in interface SlotPool

Specified by:

offerSlots in interface SlotPoolService

Parameters:

taskManagerLocation - from which the slot offers originate

taskManagerGateway - to talk to the slot offerer

offers - slot offers which are offered to the SlotPool

Returns:

A collection of accepted slot offers. The remaining slot offers are implicitly rejected.

failAllocation

public Optional<ResourceID> failAllocation(AllocationID allocationID,
                                           Exception cause)

Fail the specified allocation and release the corresponding slot if we have one. This may triggered by JobManager when some slot allocation failed with rpcTimeout. Or this could be triggered by TaskManager, when it finds out something went wrong with the slot, and decided to take it back.

Specified by:

failAllocation in interface SlotPool

Parameters:

allocationID - Represents the allocation which should be failed

cause - The cause of the failure

Returns:

Optional task executor if it has no more slots registered

failAllocation

public Optional<ResourceID> failAllocation(@Nullable
                                           ResourceID resourceId,
                                           AllocationID allocationID,
                                           Exception cause)

Description copied from interface: SlotPoolService

Fails the allocation with the given allocationId.

Specified by:

failAllocation in interface SlotPoolService

Parameters:

resourceId - taskManagerId is non-null if the signal comes from a TaskManager; if the signal comes from the ResourceManager, then it is null

allocationID - allocationId identifies which allocation to fail

cause - cause why the allocation failed

Returns:

Optional task executor if it has no more slots registered

registerTaskManager

public boolean registerTaskManager(ResourceID resourceID)

Register TaskManager to this pool, only those slots come from registered TaskManager will be considered valid. Also it provides a way for us to keep "dead" or "abnormal" TaskManagers out of this pool.

Specified by:

registerTaskManager in interface SlotPool

Specified by:

registerTaskManager in interface SlotPoolService

Parameters:

resourceID - The id of the TaskManager

Returns:

true iff a new resource id was registered

releaseTaskManager

public boolean releaseTaskManager(ResourceID resourceId,
                                  Exception cause)

Unregister TaskManager from this pool, all the related slots will be released and tasks be canceled. Called when we find some TaskManager becomes "dead" or "abnormal", and we decide to not using slots from it anymore.

Specified by:

releaseTaskManager in interface SlotPool

Specified by:

releaseTaskManager in interface SlotPoolService

Parameters:

resourceId - The id of the TaskManager

cause - for the releasing of the TaskManager

Returns:

true iff a given registered resource id was removed

createAllocatedSlotReport

public AllocatedSlotReport createAllocatedSlotReport(ResourceID taskManagerId)

Description copied from interface: SlotPool

Create report about the allocated slots belonging to the specified task manager.

Specified by:

createAllocatedSlotReport in interface SlotPool

Specified by:

createAllocatedSlotReport in interface SlotPoolService

Parameters:

taskManagerId - identifies the task manager

Returns:

the allocated slots on the task manager

timeoutPendingSlotRequest

@VisibleForTesting
protected void timeoutPendingSlotRequest(SlotRequestId slotRequestId)

checkIdleSlot

protected void checkIdleSlot()

Check the available slots, release the slot that is idle for a long time.

checkBatchSlotTimeout

protected void checkBatchSlotTimeout()

runAsync

protected void runAsync(Runnable runnable)

Execute the runnable in the main thread of the underlying RPC endpoint.

Parameters:

runnable - Runnable to be executed in the main thread of the underlying RPC endpoint

scheduleRunAsync

protected void scheduleRunAsync(Runnable runnable,
                                org.apache.flink.api.common.time.Time delay)

Execute the runnable in the main thread of the underlying RPC endpoint, with a delay of the given number of milliseconds.

Parameters:

runnable - Runnable to be executed

delay - The delay after which the runnable will be executed

scheduleRunAsync

protected void scheduleRunAsync(Runnable runnable,
                                long delay,
                                TimeUnit unit)

Execute the runnable in the main thread of the underlying RPC endpoint, with a delay of the given number of milliseconds.

Parameters:

runnable - Runnable to be executed

delay - The delay after which the runnable will be executed