Package org.apache.bookkeeper.common.util

Class Backoff

java.lang.Object

org.apache.bookkeeper.common.util.Backoff

public class Backoff
extends java.lang.Object

Implements various backoff strategies.

Strategies are defined by a Stream of durations and are intended to determine the duration after which a task is to be retried.

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
static class	Backoff.Constant	A constant backoff policy.
static class	Backoff.Exponential	A exponential backoff policy.
static class	Backoff.Jitter	A Jittered backoff policy.
static interface	Backoff.Policy	Back off policy.

Field Summary

Fields

Modifier and Type	Field	Description
static Backoff.Policy	DEFAULT

Constructor Summary

Constructors

Constructor	Description
Backoff()

Method Summary

Modifier and Type	Method	Description
static java.util.stream.Stream<java.lang.Long>	constant(long startMs)	Create a stream with constant backoffs.
static java.util.stream.Stream<java.lang.Long>	decorrelatedJittered(long startMs, long maxMs)	Create an infinite backoffs that have jitter with a random distribution between startMs and 3 times the previously selected value, capped at maxMs.
static java.util.stream.Stream<java.lang.Long>	equalJittered(long startMs, long maxMs)	Create infinite backoffs that keep half of the exponential growth, and jitter between 0 and that amount.
static java.util.stream.Stream<java.lang.Long>	exponential(long startMs, int multiplier, long maxMs)	Create a stream with exponential backoffs.
static java.util.stream.Stream<java.lang.Long>	exponentialJittered(long startMs, long maxMs)	Create a stream of exponential backoffs with jitters.

Methods inherited from class java.lang.Object

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

Field Detail

DEFAULT

public static final Backoff.Policy DEFAULT

Constructor Detail

Backoff

public Backoff()

Method Detail

constant

public static java.util.stream.Stream<java.lang.Long> constant(long startMs)

Create a stream with constant backoffs.

Parameters:

startMs - initial backoff in milliseconds

Returns:

a stream with constant backoff values.

exponential

public static java.util.stream.Stream<java.lang.Long> exponential(long startMs,
                                                                  int multiplier,
                                                                  long maxMs)

Create a stream with exponential backoffs.

Parameters:

startMs - initial backoff in milliseconds.

multiplier - the multiplier for next backoff.

maxMs - max backoff in milliseconds.

Returns:

a stream with exponential backoffs.

exponentialJittered

public static java.util.stream.Stream<java.lang.Long> exponentialJittered(long startMs,
                                                                          long maxMs)

Create a stream of exponential backoffs with jitters.

This is "full jitter" via http://www.awsarchitectureblog.com/2015/03/backoff.html

Parameters:

startMs - initial backoff in milliseconds.

maxMs - max backoff in milliseconds.

Returns:

a stream of exponential backoffs with jitters.

decorrelatedJittered

public static java.util.stream.Stream<java.lang.Long> decorrelatedJittered(long startMs,
                                                                           long maxMs)

Create an infinite backoffs that have jitter with a random distribution between startMs and 3 times the previously selected value, capped at maxMs.

this is "decorrelated jitter" via http://www.awsarchitectureblog.com/2015/03/backoff.html

Parameters:

startMs - initial backoff in milliseconds

maxMs - max backoff in milliseconds

Returns:

a stream of jitter backoffs.

equalJittered

public static java.util.stream.Stream<java.lang.Long> equalJittered(long startMs,
                                                                    long maxMs)

Create infinite backoffs that keep half of the exponential growth, and jitter between 0 and that amount.

this is "equal jitter" via http://www.awsarchitectureblog.com/2015/03/backoff.html

Parameters:

startMs - initial backoff in milliseconds.

maxMs - max backoff in milliseconds.

Returns:

a stream of exponential backoffs with jitters.