[
https://issues.apache.org/jira/browse/ARTEMIS-2131?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16653386#comment-16653386
]
Francesco Nigro edited comment on ARTEMIS-2131 at 10/17/18 11:25 AM:
---------------------------------------------------------------------
I have a strong feeling these issues are somehow related:
[https://github.com/apache/activemq-artemis/pull/2250]
If you can try that branch we can check if it fixes the issue.
The original problem seems that the buffer used while compacting is considered
a duplicate or a slice:
{code:java}
sequentialFile.writeDirect(writingChannel.toByteBuffer(), true);{code}
Due to this reason it cannot be released correctly while the mentioned PR is
making use of a call that won't try to pool or release it:
{code:java}
final ByteBuffer byteBuffer = writingChannelNioBuffer;
final int readerIndex = writingChannel.readerIndex();
byteBuffer.clear().position(readerIndex).limit(readerIndex +
writingChannel.readableBytes());
sequentialFile.blockingWriteDirect(byteBuffer, false, false);{code}
With blockingWriteDirect defined as:
{code:java}
/**
* Write directly to the file without using any intermediate buffer and wait
completion.<br>
* If {@code releaseBuffer} is {@code true} the provided {@code bytes}
should be released
* through {@link SequentialFileFactory#releaseBuffer(ByteBuffer)}, if
supported.
*
* @param bytes the ByteBuffer must be compatible with the
SequentialFile implementation (AIO or
* NIO). If {@code releaseBuffer} is {@code true} use a
buffer from
* {@link SequentialFileFactory#newBuffer(int)}, {@link
SequentialFileFactory#allocateDirectBuffer(int)}
* otherwise.
* @param sync if {@code true} will durable flush the written data
on the file, {@code false} otherwise
* @param releaseBuffer if {@code true} will release the buffer, {@code
false} otherwise
*/
void blockingWriteDirect(ByteBuffer bytes, boolean sync, boolean
releaseBuffer) throws Exception;
{code}
was (Author: [email protected]):
I have a strong feeling these issues are somehow related:
[https://github.com/apache/activemq-artemis/pull/2250]
If you can try that branch we can check if it fixes the issue.
The original problem seems that the buffer used while compacting is considered
a duplicate or a slice:
{code:java}
sequentialFile.writeDirect(writingChannel.toByteBuffer(), true);{code}
Due to this reason it cannot be released correctly while the mentioned PR is
making use of a call that won't try to pool or release it:
{code:java}
final ByteBuffer byteBuffer = writingChannelNioBuffer;
final int readerIndex = writingChannel.readerIndex();
byteBuffer.clear().position(readerIndex).limit(readerIndex +
writingChannel.readableBytes());
sequentialFile.blockingWriteDirect(byteBuffer, false, false);{code}
> Error compacting journal
> ------------------------
>
> Key: ARTEMIS-2131
> URL: https://issues.apache.org/jira/browse/ARTEMIS-2131
> Project: ActiveMQ Artemis
> Issue Type: Bug
> Affects Versions: 2.6.3
> Environment: OpenJDK 11
> Reporter: Borna
> Priority: Major
>
> Sometimes (especially when under higher load), Artemis throws the following
> exception:
> {{[10:03:01.377] ERROR [Thread-9
> (ActiveMQ-IO-server-org.apache.activemq.artemis.core.server.impl.ActiveMQServerImpl$6@2484c868)]
> o.a.a.a.journal - AMQ144003: Error compacting
> java.lang.reflect.InvocationTargetException: null at
> java.base/jdk.internal.reflect.GeneratedMethodAccessor991.invoke(Unknown
> Source) at
> java.base/jdk.internal.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
> at java.base/java.lang.reflect.Method.invoke(Method.java:566) at
> io.netty.util.internal.CleanerJava9.freeDirectBuffer(CleanerJava9.java:88) at
> io.netty.util.internal.PlatformDependent.freeDirectBuffer(PlatformDependent.java:392)
> at
> org.apache.activemq.artemis.core.io.mapped.MappedSequentialFileFactory.releaseBuffer(MappedSequentialFileFactory.java:134)
> at
> org.apache.activemq.artemis.core.io.mapped.MappedSequentialFile.writeDirect(MappedSequentialFile.java:272)
> at
> org.apache.activemq.artemis.core.io.mapped.TimedSequentialFile.writeDirect(TimedSequentialFile.java:164)
> at
> org.apache.activemq.artemis.core.journal.impl.AbstractJournalUpdateTask.flush(AbstractJournalUpdateTask.java:217)
> at
> org.apache.activemq.artemis.core.journal.impl.JournalImpl.compact(JournalImpl.java:1613)
> at
> org.apache.activemq.artemis.core.journal.impl.JournalImpl$14.run(JournalImpl.java:2159)
> at
> org.apache.activemq.artemis.utils.actors.OrderedExecutor.doTask(OrderedExecutor.java:42)
> at
> org.apache.activemq.artemis.utils.actors.OrderedExecutor.doTask(OrderedExecutor.java:31)
> at
> org.apache.activemq.artemis.utils.actors.ProcessorBase.executePendingTasks(ProcessorBase.java:66)
> at
> java.base/java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1128)
> at
> java.base/java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:628)
> at
> org.apache.activemq.artemis.utils.ActiveMQThreadFactory$1.run(ActiveMQThreadFactory.java:118)
> Caused by: java.lang.IllegalArgumentException: duplicate or slice at
> jdk.unsupported/sun.misc.Unsafe.invokeCleaner(Unsafe.java:1238) ... 17 common
> frames omitted}}
>
> As a consequence, it looks like old journal files are not reclaimed and
> consumption of disk space grows indefinitely.
> Tested using OpenJDK 11.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
