Victsm commented on a change in pull request #29855: URL: https://github.com/apache/spark/pull/29855#discussion_r500528907
########## File path: common/network-shuffle/src/main/java/org/apache/spark/network/shuffle/ErrorHandler.java ########## @@ -0,0 +1,72 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You 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 org.apache.spark.network.shuffle; + +import java.net.ConnectException; + +/** + * Plugs into {@link RetryingBlockFetcher} to further control when an exception should be retried + * and logged. + * Note: {@link RetryingBlockFetcher} will delegate the exception to this handler only when + * - remaining retries < max retries + * - exception is an IOException + */ + +public interface ErrorHandler { + boolean shouldRetryError(Throwable t); + + default boolean shouldLogError(Throwable t) { + return true; + } + + /** + * A no-op error handler instance. + */ + ErrorHandler NOOP_ERROR_HANDLER = t -> true; + + /** + * The error handler for pushing shuffle blocks to remote shuffle services. + */ + class BlockPushErrorHandler implements ErrorHandler { + + @Override + public boolean shouldRetryError(Throwable t) { + // If it is a connection time out or a connection closed exception, no need to retry. + if (t.getCause() != null && t.getCause() instanceof ConnectException) { Review comment: In our operation experience, the client usually see this exception when the shuffle service becomes unavailable and it won't come back quickly within seconds. Since push-based shuffle is best effort only, we decided to do this so we can quickly pass the shuffle service that's no longer available, instead of prolonging the push time by keeping retrying on these unavailable shuffle services which could take quite some time based on # of retries and wait time between retries. ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
