[AMQ-6067] OutOfMemoryError when expiring big amount of topic messages

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 5.10.0
• Fix Version/s: 5.14.1, 5.15.0
• Component/s: JDBC
• Labels: durable durable_subscription expiration jdbc timeToLive,
• Assignee: Christopher L. Shannon
• Reporter: Petr Havránek
• Created: 30/Nov/15 14:42
• Updated: 26/Aug/16 18:35
• Resolved: 26/Aug/16 18:34

Description

There is a problem in

org.apache.activemq.broker.region.Topic.expireMessagesTask

When there are big amount of topic messages that are going to be expired, this expireMessagesTask loads all of the messages to memory. This causes

2015-11-24 11:05:46.359 WARN  [ActiveMQ Broker[JmsEngineActivemqBroker] Scheduler] [Topic] Failed to browse Topic: test-topic
java.lang.OutOfMemoryError: Java heap space
	at oracle.sql.BLOB.getBytes(BLOB.java:204)
	at oracle.jdbc.driver.T4CBlobAccessor.getBytes(T4CBlobAccessor.java:464)
	at oracle.jdbc.driver.OracleResultSetImpl.getBytes(OracleResultSetImpl.java:676)
	at org.apache.commons.dbcp.DelegatingResultSet.getBytes(DelegatingResultSet.java:203)
	at org.apache.activemq.store.jdbc.adapter.DefaultJDBCAdapter.getBinaryData(DefaultJDBCAdapter.java:80)
	at org.apache.activemq.store.jdbc.adapter.DefaultJDBCAdapter.doRecover(DefaultJDBCAdapter.java:418)
	at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
	at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:39)
	at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:25)
	at java.lang.reflect.Method.invoke(Method.java:597)
	at org.springframework.aop.support.AopUtils.invokeJoinpointUsingReflection(AopUtils.java:309)
	at org.springframework.aop.framework.ReflectiveMethodInvocation.invokeJoinpoint(ReflectiveMethodInvocation.java:183)
	at org.springframework.aop.framework.ReflectiveMethodInvocation.proceed(ReflectiveMethodInvocation.java:150)
	at org.springframework.aop.interceptor.AbstractTraceInterceptor.invoke(AbstractTraceInterceptor.java:113)
	at org.springframework.aop.framework.ReflectiveMethodInvocation.proceed(ReflectiveMethodInvocation.java:172)
	at org.springframework.aop.framework.JdkDynamicAopProxy.invoke(JdkDynamicAopProxy.java:202)
	at $Proxy14.doRecover(Unknown Source)
	at org.apache.activemq.store.jdbc.JDBCMessageStore.recover(JDBCMessageStore.java:236)
	at org.apache.activemq.store.ProxyTopicMessageStore.recover(ProxyTopicMessageStore.java:62)
	at org.apache.activemq.broker.region.Topic.doBrowse(Topic.java:594)
	at org.apache.activemq.broker.region.Topic.access$100(Topic.java:65)
	at org.apache.activemq.broker.region.Topic$6.run(Topic.java:733)
	at org.apache.activemq.thread.SchedulerTimerTask.run(SchedulerTimerTask.java:33)
	at java.util.TimerThread.mainLoop(Timer.java:512)
	at java.util.TimerThread.run(Timer.java:462)

The problem happens when using JDBC persistency with ActiveMQ 5.10.0. After a short look to source code, the same problem could be also with 5.12.1.

Test case:

• run ActiveMQ broker with JDBC persistency
• create subscription to a topic, but do not receive the messages
• send enough number of messages with short TimeToLive
• when expireMessagesTask is scheduled, it tries to load all of the messages and causes the OutOfMemoryError

It would be fine if

org.apache.activemq.store.jdbc.JDBCMessageStore.recover(MessageRecoveryListener)

will be updated like this:

public void recover(final MessageRecoveryListener listener) throws Exception {
  // Get all the Message ids out of the database.
  TransactionContext c = persistenceAdapter.getTransactionContext();
  try {
    c = persistenceAdapter.getTransactionContext();
    adapter.doRecover(c, destination, new JDBCMessageRecoveryListener() {
      public boolean recoverMessage(long sequenceId, byte[] data) throws Exception {
        if (listener.hasSpace()) {
          Message msg = (Message)wireFormat.unmarshal(new ByteSequence(data));
          msg.getMessageId().setBrokerSequenceId(sequenceId);
          return listener.recoverMessage(msg);
        } else {
          logger.debug("Recovery limit of the messages has exceeded.");
          return false;
        }                    
      }

      public boolean recoverMessageReference(String reference) throws Exception {
        if (listener.hasSpace()) {
          return listener.recoverMessageReference(new MessageId(reference));
        } else {
          logger.debug("Recovery limit of the message references has exceeded.");
          return false;
        }
      }
    });
  } catch (SQLException e) {
    JDBCPersistenceAdapter.log("JDBC Failure: ", e);
    throw IOExceptionSupport.create("Failed to recover container. Reason: " + e, e);
  } finally {
    c.close();
  }
}

But I am not sure if this limitation is the best way, because there will be some messages that should be expired, but need to wait. So better solution might be to do this job in more separated transactions.