com.couchbase.client.core.retry.BestEffortRetryStrategy Maven / Gradle / Ivy
Show all versions of core-io Show documentation
/*
* Copyright (c) 2018 Couchbase, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package com.couchbase.client.core.retry;
import com.couchbase.client.core.msg.Request;
import com.couchbase.client.core.msg.RequestContext;
import com.couchbase.client.core.msg.Response;
import com.couchbase.client.core.retry.reactor.Backoff;
import com.couchbase.client.core.retry.reactor.IterationContext;
import java.time.Duration;
import java.util.concurrent.CompletableFuture;
/**
* Retries operations on a best-effort basis until they time out.
*
* This is the default retry strategy in the SDK and it works under the assumption that all errors which are
* retryable in the first place will be retried until the operation times out.
*
* By default, the time paused between the retries is using an exponential backoff strategy, starting at 1 milliseconds
* and up to 500 milliseconds (see {@link #DEFAULT_EXPONENTIAL_BACKOFF}). If needed, this backoff can be customized by
* instantiating a custom instance through the {@link #withExponentialBackoff(Duration, Duration, int)} method.
*
* While slightly advanced, this class is designed to be extended and the {@link #shouldRetry(Request, RetryReason)} can
* be overridden if needed to perform custom retry logic. See the method javadoc for further details.
*/
public class BestEffortRetryStrategy implements RetryStrategy {
/**
* The default backoff is an exponential backoff from 1 to 500 millis with a factor of 2.
*/
private static final Backoff DEFAULT_EXPONENTIAL_BACKOFF = Backoff.exponential(
Duration.ofMillis(1), Duration.ofMillis(500), 2, false);
/**
* Returns the default {@link BestEffortRetryStrategy} instance.
*/
public static final BestEffortRetryStrategy INSTANCE = new BestEffortRetryStrategy(DEFAULT_EXPONENTIAL_BACKOFF);
/**
* Holds the backoff delay algorithm for the retry strategy.
*/
private final Backoff backoff;
/**
* Creates a new {@link BestEffortRetryStrategy} with the {@link #DEFAULT_EXPONENTIAL_BACKOFF}.
*/
protected BestEffortRetryStrategy() {
this(DEFAULT_EXPONENTIAL_BACKOFF);
}
/**
* Creates a new {@link BestEffortRetryStrategy} with aa custom {@link Backoff}.
*
* @param backoff the custom backoff that should be used.
*/
protected BestEffortRetryStrategy(final Backoff backoff) {
this.backoff = backoff;
}
/**
* Creates a new {@link BestEffortRetryStrategy} with custom exponential backoff boundaries.
*
* @param lower the lower backoff boundary.
* @param upper the upper backoff boundary.
* @param factor the exponential factor to use.
* @return the instantiated {@link BestEffortRetryStrategy}.
*/
public static BestEffortRetryStrategy withExponentialBackoff(final Duration lower, final Duration upper,
final int factor) {
return new BestEffortRetryStrategy(Backoff.exponential(lower, upper, factor, false));
}
/**
* Determines if a request should be retried or not (and if so, after which duration).
*
* In this implementation, the operation will always be retried if it is either idempotent or if the retry reason also
* allows operations to be retried that are not idempotent. If retry is possible, the next duration based on the
* configured backoff algorithm is chosen and returned.
*
* This method is designed to be overridden by sub-classes. The common use case is that some retry reasons should
* not be retried while others should be handled as usual. In this case, override this method but make sure to
* preform custom checks for specific retry reasons (i.e. {@link RetryReason#ENDPOINT_CIRCUIT_OPEN}) but then call
* this method through super so that all the other reasons are handled properly.
*
* While most of the time a {@link CompletableFuture} is constructed immediately, it is possible to call out into
* external systems through the async mechanism without blocking all the other components in the system. If you do
* call out over the network or into files, make sure to NEVER block and follow the usual async java/reactor coding
* best practices.
*
* @param request the request that is affected.
* @param reason the reason why the operation should be retried in the first place.
* @return a future that when complete indicates the next {@link RetryAction} to take.
*/
@Override
public CompletableFuture shouldRetry(final Request request, final RetryReason reason) {
if (request.idempotent() || reason.allowsNonIdempotentRetry()) {
RequestContext ctx = request.context();
return CompletableFuture.completedFuture(RetryAction.withDuration(
backoff.apply(new RetryStrategyIterationContext(ctx.retryAttempts(), ctx.lastRetryDuration())).delay()
));
}
return CompletableFuture.completedFuture(RetryAction.noRetry());
}
@Override
public String toString() {
return "BestEffort{backoff=" + backoff + "}";
}
/**
* Helper class to make the reactor retry strategies happy we are utilizing internally to avoid code
* duplication.
*/
private static class RetryStrategyIterationContext implements IterationContext {
private final long iteration;
private final Duration lastBackoff;
RetryStrategyIterationContext(long iteration, Duration lastBackoff) {
this.iteration = iteration;
this.lastBackoff = lastBackoff;
}
@Override
public Void applicationContext() {
return null;
}
@Override
public long iteration() {
return iteration;
}
@Override
public Duration backoff() {
return lastBackoff;
}
}
}