bharathv commented on a change in pull request #3371: URL: https://github.com/apache/hbase/pull/3371#discussion_r649319325
########## File path: hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/wal/SyncFutureCache.java ########## @@ -0,0 +1,80 @@ +/* + * 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.hadoop.hbase.regionserver.wal; + +import java.util.concurrent.TimeUnit; +import org.apache.hadoop.conf.Configuration; +import org.apache.hadoop.hbase.HConstants; +import org.apache.yetus.audience.InterfaceAudience; +import org.apache.hbase.thirdparty.com.google.common.cache.Cache; +import org.apache.hbase.thirdparty.com.google.common.cache.CacheBuilder; + +/** + * A cache of {@link SyncFuture}s. This class supports two methods + * {@link SyncFutureCache#getIfPresentOrNew()} and {@link SyncFutureCache#offer()}. + * + * Usage pattern: + * SyncFuture sf = syncFutureCache.getIfPresentOrNew(); + * sf.reset(...); + * // Use the sync future + * finally: syncFutureCache.offer(sf); + * + * Offering the sync future back to the cache makes it eligible for reuse within the same thread + * context. Cache keyed by the accessing thread instance and automatically invalidated if it remains + * unused for {@link SyncFutureCache#SYNC_FUTURE_INVALIDATION_TIMEOUT_MINS} minutes. + */ +@InterfaceAudience.Private +public final class SyncFutureCache { + + private final long SYNC_FUTURE_INVALIDATION_TIMEOUT_MINS = 2; + + private final Cache<Thread, SyncFuture> syncFutureCache; + + public SyncFutureCache(final Configuration conf) { + final int handlerCount = conf.getInt(HConstants.REGION_SERVER_HANDLER_COUNT, + HConstants.DEFAULT_REGION_SERVER_HANDLER_COUNT); + syncFutureCache = CacheBuilder.newBuilder().initialCapacity(handlerCount) Review comment: Ah sweet, I heard about this Guava vs Caffeine thing, let me replace that. (didn't notice any performance issues with the patch but if we get some extra performance why not) ########## File path: hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/wal/AsyncFSWAL.java ########## @@ -263,6 +263,14 @@ public AsyncFSWAL(FileSystem fs, Abortable abortable, Path rootDir, String logDi DEFAULT_ASYNC_WAL_WAIT_ON_SHUTDOWN_IN_SECONDS); } + /** + * Helper that marks the future as DONE and offers it back to the cache. + */ + private void markFutureDoneAndOffer(SyncFuture future, long txid, Throwable t) { + future.done(txid, t); + syncFutureCache.offer(future); Review comment: @saintstack I documented the race [here](https://issues.apache.org/jira/browse/HBASE-25984?focusedCommentId=17359575&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#comment-17359575).. Once done() is called, the future can be reused immediately from another handler (without this patch). That was causing deadlocks in FSHLog. Based on my analysis of AsyncFSWAL, I think the overwrites are possible but it should not affect the correctness as the safe point is attained in a different manner. So wanted to check with Duo who is the expert on that. -- 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
