Package com.linkedin.venice.pushmonitor

Class PushStatusDecider

java.lang.Object

com.linkedin.venice.pushmonitor.PushStatusDecider

Direct Known Subclasses:

WaitAllPushStatusDecider, WaitNMinusOnePushStatusDecider

public abstract class PushStatusDecider
extends Object

Decide the offline push status by checking all replicas statuses under different offline push strategies.

Constructor Summary

Constructors

Constructor	Description
PushStatusDecider()

Method Summary

Modifier and Type	Method	Description
ExecutionStatusWithDetails	checkPushStatusAndDetailsByPartitionsStatus(OfflinePushStatus pushStatus, PartitionAssignment partitionAssignment, DisableReplicaCallback callback)	Check the current status based on PartitionStatus
protected abstract int	getNumberOfToleratedErrors()
protected ExecutionStatus	getPartitionStatus(PartitionStatus partitionStatus, int replicationFactor, Map<Instance,HelixState> instanceToStateMap, DisableReplicaCallback callback)
static List<Instance>	getReadyToServeInstances(PartitionStatus partitionStatus, PartitionAssignment partitionAssignment, int partitionId)
static List<String>	getReadyToServeInstances(Map<CharSequence,StoreReplicaStatus> replicaStatusMap)	Replicas will be considered ready to serve if their status is ExecutionStatus.COMPLETED.
static ExecutionStatus	getReplicaCurrentStatus(List<StatusSnapshot> historicStatusList)	The helper function is used by both controller and router (leader/follower stores); please be cautious when modifying it because it would affect both components.
abstract OfflinePushStrategy	getStrategy()
Optional<String>	hasEnoughNodesToStartPush(String kafkaTopic, int replicationFactor, ResourceAssignment resourceAssignment, Optional<String> previousReason)
protected abstract boolean	hasEnoughReplicasForOnePartition(int actual, int expected)

Methods inherited from class java.lang.Object

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

Constructor Details

PushStatusDecider

public PushStatusDecider()

Method Details

checkPushStatusAndDetailsByPartitionsStatus

public ExecutionStatusWithDetails checkPushStatusAndDetailsByPartitionsStatus(OfflinePushStatus pushStatus,
 PartitionAssignment partitionAssignment,
 DisableReplicaCallback callback)

Check the current status based on PartitionStatus

getReadyToServeInstances

public static List<Instance> getReadyToServeInstances(PartitionStatus partitionStatus,
 PartitionAssignment partitionAssignment,
 int partitionId)

getReadyToServeInstances

public static List<String> getReadyToServeInstances(Map<CharSequence,StoreReplicaStatus> replicaStatusMap)

Replicas will be considered ready to serve if their status is ExecutionStatus.COMPLETED. More information is needed if you'd like to change/support other behaviors such as routing to the leader replica.

Parameters:

replicaStatusMap -

Returns:

List of ready to serve instance ids

hasEnoughNodesToStartPush

public Optional<String> hasEnoughNodesToStartPush(String kafkaTopic,
 int replicationFactor,
 ResourceAssignment resourceAssignment,
 Optional<String> previousReason)

Returns:

status details: if empty, push has enough replicas in every partition, otherwise, message contains details

getStrategy

public abstract OfflinePushStrategy getStrategy()

hasEnoughReplicasForOnePartition

protected abstract boolean hasEnoughReplicasForOnePartition(int actual,
 int expected)

getNumberOfToleratedErrors

protected abstract int getNumberOfToleratedErrors()

getPartitionStatus

protected ExecutionStatus getPartitionStatus(PartitionStatus partitionStatus,
 int replicationFactor,
 Map<Instance,HelixState> instanceToStateMap,
 DisableReplicaCallback callback)

getReplicaCurrentStatus

public static ExecutionStatus getReplicaCurrentStatus(List<StatusSnapshot> historicStatusList)

The helper function is used by both controller and router (leader/follower stores); please be cautious when modifying it because it would affect both components.