[
https://issues.apache.org/jira/browse/FLUME-2067?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13680100#comment-13680100
]
Edward Sargisson commented on FLUME-2067:
-----------------------------------------
I have been doing some debugging on this issue.
My current conclusion is that having the Avro tranceiver and BerkeleyDB in the
same thread does not work together. Avro uses some utilities from
java.util.concurrent, such as a CountDownLatch, which will throw
InterruptedException. BerkeleyDB, at certain points, explicitly checks to see
if its thread has been interrupted and throws an exception if it has. My
evidence for this is to run the code in a debugger that breaks on
InterruptedException. I commonly get breaks in java.util.concurrent code.
The repro for this is to create a succeeding Flume node which fills ups its
channel (in my case, it runs out of memory and stops forwarding messages on).
> Logging from log4j2 FlumeAppender from Jetty webapp to Avro source with full
> queue raises ClosedByInterruptException or IllegalStateException: Channel
> closed
> --------------------------------------------------------------------------------------------------------------------------------------------------------------
>
> Key: FLUME-2067
> URL: https://issues.apache.org/jira/browse/FLUME-2067
> Project: Flume
> Issue Type: Bug
> Components: File Channel, Sinks+Sources
> Affects Versions: v1.3.1
> Reporter: Edward Sargisson
> Attachments: flume-embedded-web-flume-2067.tar.gz
>
>
> Attempting to embed a Flume agent in another app does not work very well. I
> have found a repro of a very simple Jetty app using the log4j2 FlumeAppender
> to connect to a subsequent Flume agent with a full channel.
> The signature of this problem also occurs when I have a custom version of
> Flume with log4j2 trying to log its own logs via Flume. In that case, if I
> have two servers configured to log to each other then neither will start
> correctly because of issues pushing events to the subsequent agent.
> Note that this is documented against Flume 1.3.1 because that's what the
> log4j2 Flume Appender uses. Flume 1.4.0 changes the classes sufficiently that
> it is non-trivial to attempt a repro.
> Note that I include some suppositions at the end of this work item. I haven't
> been able to properly determine what is causing the failure so you may not
> want to read them until you've done your own research.
> Steps:
> 1. Setup an additional Flume server (the subsequent server) with an avro
> source and make the channel fill up.
> 2. Extract the enclosed project. Edit the
> flume-embedded-hot-deploy/src/main/resource/log4j2.xml and configure the
> Agent for the FlumeAppender with the details of the subsequent server.
> 3. mvn clean install
> 4. Change to flume-embedded-hot-deploy
> 5. mvn clean package -P debug (note that you can set it to suspend until a
> debugger is attached with mvn clean package -P debug,suspend)
> 6. Wait for Jetty to startup - and then for a few seconds.
> Expected results:
> Some complaints about the subsequent server being full but an otherwise happy
> server.
> Actual results:
> When using the log4j2 Persistent agent (which uses Berkeley DB as a store):
> 2013-06-03 14:01:14,804 INFO [main] server.AbstractConnector
> (AbstractConnector.java:265) - Started
> ServerConnector@75a213c0{HTTP/1.1}{0.0.0.0:8080}
> 2013-06-03 14:01:22,779 DEBUG [Thread-3] ipc.NettyTransceiver
> (NettyTransceiver.java:314) - Disconnecting from
> collector1-sal-flex-van.dev-globalrelay.net/10.21.30.20:36892
> 2013-06-03 14:01:22,789 ERROR An exception occurred processing Appender
> FlumeAppender org.apache.logging.log4j.LoggingException: Exception occurred
> writing log event
> at
> org.apache.logging.log4j.flume.appender.FlumePersistentManager.send(FlumePersistentManager.java:176)
> at
> org.apache.logging.log4j.flume.appender.FlumeAppender.append(FlumeAppender.java:86)
> at
> org.apache.logging.log4j.core.config.AppenderControl.callAppender(AppenderControl.java:102)
> at
> org.apache.logging.log4j.core.config.LoggerConfig.callAppenders(LoggerConfig.java:424)
> at
> org.apache.logging.log4j.core.config.LoggerConfig.log(LoggerConfig.java:405)
> at
> org.apache.logging.log4j.core.config.LoggerConfig.log(LoggerConfig.java:366)
> at org.apache.logging.log4j.core.Logger.log(Logger.java:110)
> at
> org.apache.logging.log4j.spi.AbstractLoggerWrapper.log(AbstractLoggerWrapper.java:55)
> at org.slf4j.impl.SLF4JLogger.debug(SLF4JLogger.java:139)
> at
> org.apache.avro.ipc.NettyTransceiver$NettyClientAvroHandler.handleUpstream(NettyTransceiver.java:491)
> at
> org.jboss.netty.channel.DefaultChannelPipeline.sendUpstream(DefaultChannelPipeline.java:564)
> at
> org.jboss.netty.channel.DefaultChannelPipeline$DefaultChannelHandlerContext.sendUpstream(DefaultChannelPipeline.java:792)
> at
> org.jboss.netty.handler.codec.frame.FrameDecoder.cleanup(FrameDecoder.java:348)
> at
> org.jboss.netty.handler.codec.frame.FrameDecoder.channelDisconnected(FrameDecoder.java:230)
> at
> org.jboss.netty.channel.SimpleChannelUpstreamHandler.handleUpstream(SimpleChannelUpstreamHandler.java:107)
> at
> org.jboss.netty.channel.DefaultChannelPipeline.sendUpstream(DefaultChannelPipeline.java:564)
> at
> org.jboss.netty.channel.DefaultChannelPipeline.sendUpstream(DefaultChannelPipeline.java:559)
> at
> org.jboss.netty.channel.Channels.fireChannelDisconnected(Channels.java:399)
> at org.jboss.netty.channel.Channels$4.run(Channels.java:389)
> at
> org.jboss.netty.channel.socket.ChannelRunnableWrapper.run(ChannelRunnableWrapper.java:41)
> at
> org.jboss.netty.channel.socket.nio.AbstractNioWorker.processEventQueue(AbstractNioWorker.java:352)
> at
> org.jboss.netty.channel.socket.nio.AbstractNioWorker.run(AbstractNioWorker.java:236)
> at org.jboss.netty.channel.socket.nio.NioWorker.run(NioWorker.java:38)
> at
> org.jboss.netty.util.internal.DeadLockProofWorker$1.run(DeadLockProofWorker.java:42)
> at
> java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145)
> at
> java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615)
> at java.lang.Thread.run(Thread.java:722)
> Caused by: com.sleepycat.je.ThreadInterruptedException: (JE 5.0.73)
> Environment must be closed, caused by:
> com.sleepycat.je.ThreadInterruptedException: Environment invalid because of
> previous exception: (JE 5.0.73) /var/local/flume/castellan-reader-berkeley-db
> Channel closed, may be due to thread interrupt THREAD_INTERRUPTED:
> InterruptedException may cause incorrect internal state, unable to continue.
> Environment is invalid and must be closed.
> at
> com.sleepycat.je.ThreadInterruptedException.wrapSelf(ThreadInterruptedException.java:99)
> at
> com.sleepycat.je.dbi.EnvironmentImpl.checkIfInvalid(EnvironmentImpl.java:1512)
> at com.sleepycat.je.Transaction.checkEnv(Transaction.java:850)
> at com.sleepycat.je.Transaction.abort(Transaction.java:204)
> at
> org.apache.logging.log4j.flume.appender.FlumePersistentManager.send(FlumePersistentManager.java:171)
> ... 26 more
> Caused by: com.sleepycat.je.ThreadInterruptedException: Environment invalid
> because of previous exception: (JE 5.0.73)
> /var/local/flume/castellan-reader-berkeley-db Channel closed, may be due to
> thread interrupt THREAD_INTERRUPTED: InterruptedException may cause incorrect
> internal state, unable to continue. Environment is invalid and must be closed.
> at
> com.sleepycat.je.log.FileManager$LogEndFileDescriptor.force(FileManager.java:3054)
> at
> com.sleepycat.je.log.FileManager$LogEndFileDescriptor.access$500(FileManager.java:2710)
> at com.sleepycat.je.log.FileManager.syncLogEnd(FileManager.java:2022)
> at com.sleepycat.je.log.FSyncManager.executeFSync(FSyncManager.java:282)
> at com.sleepycat.je.log.FSyncManager.fsync(FSyncManager.java:233)
> at com.sleepycat.je.log.FileManager.groupSync(FileManager.java:2070)
> at com.sleepycat.je.log.LogManager.multiLog(LogManager.java:403)
> at com.sleepycat.je.log.LogManager.log(LogManager.java:335)
> at com.sleepycat.je.txn.Txn.logCommitEntry(Txn.java:957)
> at com.sleepycat.je.txn.Txn.commit(Txn.java:719)
> at com.sleepycat.je.txn.Txn.commit(Txn.java:584)
> at com.sleepycat.je.Transaction.commit(Transaction.java:317)
> at
> org.apache.logging.log4j.flume.appender.FlumePersistentManager.send(FlumePersistentManager.java:167)
> ... 26 more
> Caused by: java.nio.channels.ClosedByInterruptException
> at
> java.nio.channels.spi.AbstractInterruptibleChannel.end(AbstractInterruptibleChannel.java:202)
> at sun.nio.ch.FileChannelImpl.force(FileChannelImpl.java:367)
> at
> com.sleepycat.je.log.FileManager$LogEndFileDescriptor.force(FileManager.java:3043)
> ... 38 more
> When using log4j2 embedded agent (which is a Flume node running inside the
> app):
> 2013-06-03 14:14:28,207 DEBUG Calling createLoggers on class
> org.apache.logging.log4j.core.config.plugins.LoggersPlugin for element
> loggers with params(loggers={org.apache.log4j.xml, org.springframework,
> org.eclipse.jetty, org.elasticsearch, human.com.globalrelay,
> machine.com.globalrelay, root})
> 2013-06-03 14:14:28,212 DEBUG Reconfiguration completed
> 2013-06-03 14:14:28,244 ERROR An exception occurred processing Appender
> FlumeAppender java.lang.IllegalStateException: Channel closed
> [channel=primary]
> at
> org.apache.flume.channel.file.FileChannel.createTransaction(FileChannel.java:355)
> at
> org.apache.flume.channel.BasicChannelSemantics.getTransaction(BasicChannelSemantics.java:122)
> at
> org.apache.flume.channel.ChannelProcessor.processEvent(ChannelProcessor.java:260)
> at
> org.apache.logging.log4j.flume.appender.Log4jEventSource.send(Log4jEventSource.java:59)
> at
> org.apache.logging.log4j.flume.appender.FlumeEmbeddedManager.send(FlumeEmbeddedManager.java:123)
> at
> org.apache.logging.log4j.flume.appender.FlumeAppender.append(FlumeAppender.java:86)
> at
> org.apache.logging.log4j.core.config.AppenderControl.callAppender(AppenderControl.java:102)
> at
> org.apache.logging.log4j.core.config.LoggerConfig.callAppenders(LoggerConfig.java:424)
> at
> org.apache.logging.log4j.core.config.LoggerConfig.log(LoggerConfig.java:405)
> at
> org.apache.logging.log4j.core.config.LoggerConfig.log(LoggerConfig.java:366)
> at org.apache.logging.log4j.core.Logger.log(Logger.java:110)
> at
> org.apache.logging.log4j.spi.AbstractLoggerWrapper.log(AbstractLoggerWrapper.java:55)
> at org.slf4j.impl.SLF4JLogger.debug(SLF4JLogger.java:139)
> at
> org.apache.avro.ipc.NettyTransceiver.<init>(NettyTransceiver.java:195)
> at
> org.apache.avro.ipc.NettyTransceiver.<init>(NettyTransceiver.java:152)
> at
> org.apache.flume.api.NettyAvroRpcClient.connect(NettyAvroRpcClient.java:106)
> at
> org.apache.flume.api.NettyAvroRpcClient.connect(NettyAvroRpcClient.java:93)
> at
> org.apache.flume.api.NettyAvroRpcClient.configure(NettyAvroRpcClient.java:507)
> at
> org.apache.flume.api.RpcClientFactory.getInstance(RpcClientFactory.java:88)
> at org.apache.flume.sink.AvroSink.createConnection(AvroSink.java:182)
> at org.apache.flume.sink.AvroSink.start(AvroSink.java:242)
> at
> org.apache.flume.sink.AbstractSinkProcessor.start(AbstractSinkProcessor.java:41)
> at org.apache.flume.SinkRunner.start(SinkRunner.java:79)
> at
> org.apache.flume.lifecycle.LifecycleSupervisor$MonitorRunnable.run(LifecycleSupervisor.java:236)
> at
> java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:471)
> at
> java.util.concurrent.FutureTask$Sync.innerRunAndReset(FutureTask.java:351)
> at java.util.concurrent.FutureTask.runAndReset(FutureTask.java:178)
> at
> java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask.access$301(ScheduledThreadPoolExecutor.java:178)
> at
> java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask.run(ScheduledThreadPoolExecutor.java:293)
> at
> java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145)
> at
> java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615)
> at java.lang.Thread.run(Thread.java:722)
> Suppositions:
> I believe this is an issue with the handling of a failed put or take. The
> failure path interacts with rollback or the Avro client in bad ways.
> For example, for the Berkeley DB case, the Avro client uses a
> SynchronousQueue to do rendezvous. That queue uses InterruptedException
> internally. However, the FileChannel uses the NIO FileChannel (and
> AbstractInterruptibleChannel) which fails if its thread gets interrupted.
> Some docs:
> http://jira.codehaus.org/browse/JETTY-80
> http://www.oracle.com/technetwork/products/berkeleydb/if-097768.html
> For the embedded Flume case I haven't been able to establish the cause. I
> think there's a double rollback going on - or possibly the reading thread and
> the writing thread are being interfered with by other threads. (My initial
> theory was that Flume logging itself got tied up with singleton type issues
> but the Jetty repro I just found challenges that).
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
