carterkozak commented on a change in pull request #452: URL: https://github.com/apache/logging-log4j2/pull/452#discussion_r554036911
########## File path: log4j-core/src/main/java/org/apache/logging/log4j/core/appender/AsyncAppenderEventDispatcher.java ########## @@ -0,0 +1,172 @@ +/* + * 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.logging.log4j.core.appender; + +import org.apache.logging.log4j.Logger; +import org.apache.logging.log4j.core.LogEvent; +import org.apache.logging.log4j.core.config.AppenderControl; +import org.apache.logging.log4j.core.impl.Log4jLogEvent; +import org.apache.logging.log4j.core.util.Log4jThread; +import org.apache.logging.log4j.status.StatusLogger; + +import java.util.List; +import java.util.concurrent.BlockingQueue; +import java.util.concurrent.atomic.AtomicBoolean; +import java.util.concurrent.atomic.AtomicLong; + +class AsyncAppenderEventDispatcher extends Log4jThread { + + private static final LogEvent STOP_EVENT = new Log4jLogEvent(); + + private static final AtomicLong THREAD_COUNTER = new AtomicLong(0); + + private static final Logger LOGGER = StatusLogger.getLogger(); + + private final AppenderControl errorAppender; + + private final List<AppenderControl> appenders; + + private final BlockingQueue<LogEvent> queue; + + private final AtomicBoolean stoppedRef; + + AsyncAppenderEventDispatcher( + final String name, + final AppenderControl errorAppender, + final List<AppenderControl> appenders, + final BlockingQueue<LogEvent> queue) { + super("AsyncAppenderEventDispatcher-" + THREAD_COUNTER.incrementAndGet() + "-" + name); + this.errorAppender = errorAppender; + this.appenders = appenders; + this.queue = queue; + this.stoppedRef = new AtomicBoolean(false); + } + + @Override + public void run() { + LOGGER.trace("{} has started.", getName()); + dispatchAll(); + dispatchRemaining(); + } + + private void dispatchAll() { + while (!stoppedRef.get()) { + LogEvent event; + try { + event = queue.take(); + } catch (final InterruptedException ignored) { + // Restore the interrupted flag cleared when the exception is caught. + interrupt(); + break; + } + if (event == STOP_EVENT) { + break; + } + event.setEndOfBatch(queue.isEmpty()); + dispatch(event); + } + LOGGER.trace("{} has stopped.", getName()); + } + + private void dispatchRemaining() { + int eventCount = 0; + while (true) { + // Note the non-blocking Queue#poll() method! + final LogEvent event = queue.poll(); + if (event == null || event == STOP_EVENT) { Review comment: > Shouldn't we simply discard anything that is sent after the stop() signal? I think this depends how we define "after" and the "stop signal" ;-) Consider: Thread T1 calls asyncAppender.append with event E T1 checks if the appender is running, this succeeds Thread T2 calls asyncAppender.stop() T2 sets stopped = true T2 submits the STOP_EVENT Thread T1 submits event E to the queue An event is submitted prior to stop, but the STOP_EVENT is received first. Ideally if the event is ignored, we would provide a status message describing that the event was not recorded because the appender was stopped -- events shouldn't disappear, and ideally we don't leave objects (with references to potentially large user-provided parameters) sitting in the queue on heap. That said, continuing to read from the queue after STOP_EVENT doesn't guarantee that we solve the problem (offer may not have been invoked yet!), only makes it less likely that we encounter the race. ---------------------------------------------------------------- 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]
