jihoonson commented on a change in pull request #10082: URL: https://github.com/apache/druid/pull/10082#discussion_r448136101
########## File path: server/src/main/java/org/apache/druid/query/RetryQueryRunner.java ########## @@ -0,0 +1,250 @@ +/* + * 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.druid.query; + +import com.fasterxml.jackson.core.type.TypeReference; +import com.fasterxml.jackson.databind.ObjectMapper; +import com.google.common.annotations.VisibleForTesting; +import com.google.common.base.Preconditions; +import org.apache.druid.java.util.common.ISE; +import org.apache.druid.java.util.common.guava.BaseSequence; +import org.apache.druid.java.util.common.guava.BaseSequence.IteratorMaker; +import org.apache.druid.java.util.common.guava.MergeSequence; +import org.apache.druid.java.util.common.guava.Sequence; +import org.apache.druid.java.util.common.guava.Yielder; +import org.apache.druid.java.util.common.guava.YieldingAccumulator; +import org.apache.druid.java.util.common.guava.YieldingSequenceBase; +import org.apache.druid.java.util.common.logger.Logger; +import org.apache.druid.query.context.ResponseContext; +import org.apache.druid.query.context.ResponseContext.Key; +import org.apache.druid.segment.SegmentMissingException; + +import java.util.ArrayList; +import java.util.Collections; +import java.util.Iterator; +import java.util.List; +import java.util.NoSuchElementException; +import java.util.concurrent.ConcurrentHashMap; +import java.util.function.BiFunction; + +public class RetryQueryRunner<T> implements QueryRunner<T> +{ + private static final Logger LOG = new Logger(RetryQueryRunner.class); + + private final QueryRunner<T> baseRunner; + private final BiFunction<Query<T>, List<SegmentDescriptor>, QueryRunner<T>> retryRunnerCreateFn; + private final RetryQueryRunnerConfig config; + private final ObjectMapper jsonMapper; + + /** + * Runnable executed after the broker creates query distribution tree for the first attempt. This is only + * for testing and must not be used in production code. + */ + private final Runnable runnableAfterFirstAttempt; + + private int totalNumRetries; + + public RetryQueryRunner( + QueryRunner<T> baseRunner, + BiFunction<Query<T>, List<SegmentDescriptor>, QueryRunner<T>> retryRunnerCreateFn, + RetryQueryRunnerConfig config, + ObjectMapper jsonMapper + ) + { + this(baseRunner, retryRunnerCreateFn, config, jsonMapper, () -> {}); + } + + /** + * Constructor only for testing. + */ + @VisibleForTesting + RetryQueryRunner( + QueryRunner<T> baseRunner, + BiFunction<Query<T>, List<SegmentDescriptor>, QueryRunner<T>> retryRunnerCreateFn, + RetryQueryRunnerConfig config, + ObjectMapper jsonMapper, + Runnable runnableAfterFirstAttempt + ) + { + this.baseRunner = baseRunner; + this.retryRunnerCreateFn = retryRunnerCreateFn; + this.config = config; + this.jsonMapper = jsonMapper; + this.runnableAfterFirstAttempt = runnableAfterFirstAttempt; + } + + @VisibleForTesting + int getTotalNumRetries() + { + return totalNumRetries; + } + + @Override + public Sequence<T> run(final QueryPlus<T> queryPlus, final ResponseContext context) + { + return new YieldingSequenceBase<T>() + { + @Override + public <OutType> Yielder<OutType> toYielder(OutType initValue, YieldingAccumulator<OutType, T> accumulator) + { + final Sequence<Sequence<T>> retryingSequence = new BaseSequence<>( + new IteratorMaker<Sequence<T>, RetryingSequenceIterator>() + { + @Override + public RetryingSequenceIterator make() + { + return new RetryingSequenceIterator(queryPlus, context, baseRunner, runnableAfterFirstAttempt); + } + + @Override + public void cleanup(RetryingSequenceIterator iterFromMake) + { + totalNumRetries = iterFromMake.retryCount; + } + } + ); + return new MergeSequence<>(queryPlus.getQuery().getResultOrdering(), retryingSequence) + .toYielder(initValue, accumulator); + } + }; + } + + private List<SegmentDescriptor> getMissingSegments(QueryPlus<T> queryPlus, final ResponseContext context) Review comment: Hmm, I was aware of this issue, but ignored since the max size of the serialized responseContext is 7k. The missingSegments is a list of `SegmentDescriptor` which has an `interval`, a `version`, and a `partitionNumber`. Since its serialized size is about 80 bytes, each historical can report up to about 90 missing segments. I thought it wouldn't happen often, but well, I guess it depends on how your cluster is configured and how many segments the query reads. > I'm not sure what better we can really do without a fairly large refactoring involving making the broker keep track of which segments went to which direct druid client/sequence, and allowing historicals to basically give up if missing so many that the list would be truncated, returning an empty sequence and an indicator that **all** of the segments it was queried for should be retried. Hmm, I was thinking that we don't have to keep track of all segments, but can send a flag in the query context to historicals which makes the query failed immediately if the serialized responseContext is truncated. The broker will catch the failure and cancel the entire query. This won't retry on the truncated serialization, but I think it makes sense since it's a sort of failure of the retry system. ---------------------------------------------------------------- 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: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: commits-unsubscr...@druid.apache.org For additional commands, e-mail: commits-h...@druid.apache.org