activemq-artemis/docs/user-manual/en/large-messages.xml

297 lines
16 KiB
XML

<?xml version="1.0" encoding="UTF-8"?>
<!-- ============================================================================= -->
<!-- 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. -->
<!-- ============================================================================= -->
<!DOCTYPE chapter PUBLIC "-//OASIS//DTD DocBook XML V4.5//EN" "http://www.oasis-open.org/docbook/xml/4.5/docbookx.dtd" [
<!ENTITY % BOOK_ENTITIES SYSTEM "ActiveMQ_User_Manual.ent">
%BOOK_ENTITIES;
]>
<chapter id="large-messages">
<title>Large Messages</title>
<para>ActiveMQ supports sending and receiving of huge messages, even when the client and server
are running with limited memory. The only realistic limit to the size of a message that can
be sent or consumed is the amount of disk space you have available. We have tested sending
and consuming messages up to 8 GiB in size with a client and server running in just 50MiB of
RAM!</para>
<para>To send a large message, the user can set an <literal>InputStream</literal> on a message
body, and when that message is sent, ActiveMQ will read the <literal>InputStream</literal>. A
<literal>FileInputStream</literal> could be used for example to send a huge message from
a huge file on disk.</para>
<para>As the <literal>InputStream</literal> is read the data is sent to the server as a stream
of fragments. The server persists these fragments to disk as it receives them and when the
time comes to deliver them to a consumer they are read back of the disk, also in fragments
and sent down the wire. When the consumer receives a large message it initially receives
just the message with an empty body, it can then set an <literal>OutputStream</literal> on
the message to stream the huge message body to a file on disk or elsewhere. At no time is
the entire message body stored fully in memory, either on the client or the server.</para>
<section id="large.message.configuring">
<title>Configuring the server</title>
<para>Large messages are stored on a disk directory on the server side, as configured on the
main configuration file.</para>
<para>The configuration property <literal>large-messages-directory</literal> specifies where
large messages are stored.</para>
<programlisting>
&lt;configuration xmlns="urn:activemq"
xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
xsi:schemaLocation="urn:activemq /schema/activemq-configuration.xsd">
...
&lt;large-messages-directory>/data/large-messages&lt;/large-messages-directory>
...
&lt;/configuration</programlisting>
<para>By default the large message directory is <literal>data/largemessages</literal></para>
<para>For the best performance we recommend large messages directory is stored on a
different physical volume to the message journal or paging directory.</para>
</section>
<section>
<title>Configuring Parameters</title>
<para>Any message larger than a certain size is considered a large message. Large messages
will be split up and sent in fragments. This is determined by the parameter <literal
>min-large-message-size</literal></para>
<note>
<para>ActiveMQ messages are encoded using 2 bytes per character so if the message data is filled
with ASCII characters (which are 1 byte) the size of the resulting ActiveMQ message would roughly
double. This is important when calculating the size of a "large" message as it may appear to be
less than the <literal>min-large-message-size</literal> before it is sent, but it then turns into
a "large" message once it is encoded.</para>
</note>
<para>The default value is 100KiB.</para>
<section id="large-messages.core.config">
<title>Using Core API</title>
<para>If the ActiveMQ Core API is used, the minimal large message size is specified by
<literal>ServerLocator.setMinLargeMessageSize</literal>.</para>
<programlisting>
ServerLocator locator = ActiveMQClient.createServerLocatorWithoutHA(new TransportConfiguration(NettyConnectorFactory.class.getName()))
locator.setMinLargeMessageSize(25 * 1024);
ClientSessionFactory factory = ActiveMQClient.createClientSessionFactory();</programlisting>
<para><xref linkend="configuring-transports.client.side"/> will provide more information
on how to instantiate the session factory.</para>
</section>
<section>
<title>Using JMS</title>
<para>If JNDI is used to look up the connection factory, the minimum large message size
is specified in <literal>activemq-jms.xml</literal></para>
<programlisting>...
&lt;connection-factory name="ConnectionFactory">
&lt;connectors>
&lt;connector-ref connector-name="netty"/>
&lt;/connectors>
&lt;entries>
&lt;entry name="ConnectionFactory"/>
&lt;entry name="XAConnectionFactory"/>
&lt;/entries>
&lt;min-large-message-size>250000&lt;/min-large-message-size>
&lt;/connection-factory>
...</programlisting>
<para>If the connection factory is being instantiated directly, the minimum large
message size is specified by <literal
>ActiveMQConnectionFactory.setMinLargeMessageSize</literal>.</para>
</section>
<section>
<title>Compressed Large Messages</title>
<para>
You can choose to send large messages in compressed form using <literal>
compress-large-messages</literal> attributes.
</para>
<section>
<title><literal>compress-large-messages</literal></title>
<para>If you specify the boolean property <literal>compress-large-messages</literal> on
the <literal>server locator</literal> or <literal>ConnectionFactory</literal> as true, The
system will use the ZIP algorithm to compress the message body as the message is
transferred to the server's side. Notice that there's no special treatment at the
server's side, all the compressing and uncompressing is done at the client.</para>
<para>If the compressed size of a large message is below <literal>
min-large-message-size</literal>, it is sent to server as regular messages. This means
that the message won't be written into the server's large-message
data directory, thus reducing the disk I/O.</para>
</section>
<section>
<para>If you use JMS, you can achieve large messages compression by configuring your
connection factories. For example,</para>
<programlisting>...
&lt;connection-factory name="ConnectionFactory">
&lt;connectors>
&lt;connector-ref connector-name="netty"/>
&lt;/connectors>
...
&lt;compress-large-messages>true&lt;/compress-large-messages>
&lt;/connection-factory>
...</programlisting>
</section>
</section>
</section>
<section>
<title>Streaming large messages</title>
<para>ActiveMQ supports setting the body of messages using input and output streams (<literal
>java.lang.io</literal>)</para>
<para>These streams are then used directly for sending (input streams) and receiving (output
streams) messages.</para>
<para>When receiving messages there are 2 ways to deal with the output stream; you may
choose to block while the output stream is recovered using the method <literal
>ClientMessage.saveOutputStream</literal> or alternatively using the method <literal
>ClientMessage.setOutputstream</literal> which will asynchronously write the message
to the stream. If you choose the latter the consumer must be kept alive until the
message has been fully received.</para>
<para>You can use any kind of stream you like. The most common use case is to send files
stored in your disk, but you could also send things like JDBC Blobs, <literal
>SocketInputStream</literal>, things you recovered from <literal
>HTTPRequests</literal> etc. Anything as long as it implements <literal
>java.io.InputStream</literal> for sending messages or <literal
>java.io.OutputStream</literal> for receiving them.</para>
<section>
<title>Streaming over Core API</title>
<para>The following table shows a list of methods available at <literal
>ClientMessage</literal> which are also available through JMS by the use of
object properties.</para>
<table frame="topbot" id="large-messages.ClientMessageAPI">
<title>org.apache.activemq.api.core.client.ClientMessage API</title>
<tgroup cols="3">
<colspec colname="Name" colnum="1"/>
<colspec colname="Descr" colnum="2"/>
<colspec colname="JMS" colnum="3"/>
<thead>
<row>
<entry>Name</entry>
<entry>Description</entry>
<entry>JMS Equivalent Property</entry>
</row>
</thead>
<tbody>
<row>
<entry>setBodyInputStream(InputStream)</entry>
<entry>Set the InputStream used to read a message body when sending
it.</entry>
<entry>JMS_HQ_InputStream</entry>
</row>
<row>
<entry>setOutputStream(OutputStream)</entry>
<entry>Set the OutputStream that will receive the body of a message.
This method does not block.</entry>
<entry>JMS_HQ_OutputStream</entry>
</row>
<row>
<entry>saveOutputStream(OutputStream)</entry>
<entry>Save the body of the message to the <literal
>OutputStream</literal>. It will block until the entire content
is transferred to the <literal>OutputStream</literal>.</entry>
<entry>JMS_HQ_SaveStream</entry>
</row>
</tbody>
</tgroup>
</table>
<para>To set the output stream when receiving a core message: </para>
<programlisting>
...
ClientMessage msg = consumer.receive(...);
// This will block here until the stream was transferred
msg.saveOutputStream(someOutputStream);
ClientMessage msg2 = consumer.receive(...);
// This will not wait the transfer to finish
msg.setOutputStream(someOtherOutputStream);
...</programlisting>
<para> Set the input stream when sending a core message: </para>
<programlisting>
...
ClientMessage msg = session.createMessage();
msg.setInputStream(dataInputStream);
...</programlisting>
<para>Notice also that for messages with more than 2GiB the getBodySize() will return
invalid values since this is an integer (which is also exposed to the JMS API). On
those cases you can use the message property _HQ_LARGE_SIZE.</para>
</section>
<section id="large-messages.streaming.over.jms">
<title>Streaming over JMS</title>
<para>When using JMS, ActiveMQ maps the streaming methods on the core API (see <xref
linkend="large-messages.ClientMessageAPI"/>) by setting object properties . You
can use the method <literal>Message.setObjectProperty</literal> to set the input and
output streams.</para>
<para>The <literal>InputStream</literal> can be defined through the JMS Object Property
JMS_HQ_InputStream on messages being sent:</para>
<programlisting>
BytesMessage message = session.createBytesMessage();
FileInputStream fileInputStream = new FileInputStream(fileInput);
BufferedInputStream bufferedInput = new BufferedInputStream(fileInputStream);
message.setObjectProperty("JMS_HQ_InputStream", bufferedInput);
someProducer.send(message);</programlisting>
<para>The <literal>OutputStream</literal> can be set through the JMS Object Property
JMS_HQ_SaveStream on messages being received in a blocking way.</para>
<programlisting>
BytesMessage messageReceived = (BytesMessage)messageConsumer.receive(120000);
File outputFile = new File("huge_message_received.dat");
FileOutputStream fileOutputStream = new FileOutputStream(outputFile);
BufferedOutputStream bufferedOutput = new BufferedOutputStream(fileOutputStream);
// This will block until the entire content is saved on disk
messageReceived.setObjectProperty("JMS_HQ_SaveStream", bufferedOutput);</programlisting>
<para>Setting the <literal>OutputStream</literal> could also be done in a non blocking
way using the property JMS_HQ_OutputStream.</para>
<programlisting>
// This won't wait the stream to finish. You need to keep the consumer active.
messageReceived.setObjectProperty("JMS_HQ_OutputStream", bufferedOutput);</programlisting>
<note>
<para>When using JMS, Streaming large messages are only supported on <literal
>StreamMessage</literal> and <literal>BytesMessage</literal>.</para>
</note>
</section>
</section>
<section>
<title>Streaming Alternative</title>
<para>If you choose not to use the <literal>InputStream</literal> or <literal
>OutputStream</literal> capability of ActiveMQ You could still access the data
directly in an alternative fashion.</para>
<para>On the Core API just get the bytes of the body as you normally would.</para>
<programlisting>
ClientMessage msg = consumer.receive();
byte[] bytes = new byte[1024];
for (int i = 0 ; i &lt; msg.getBodySize(); i += bytes.length)
{
msg.getBody().readBytes(bytes);
// Whatever you want to do with the bytes
}</programlisting>
<para>If using JMS API, <literal>BytesMessage</literal> and <literal>StreamMessage</literal>
also supports it transparently.</para>
<programlisting>
BytesMessage rm = (BytesMessage)cons.receive(10000);
byte data[] = new byte[1024];
for (int i = 0; i &lt; rm.getBodyLength(); i += 1024)
{
int numberOfBytes = rm.readBytes(data);
// Do whatever you want with the data
} </programlisting>
</section>
<section id="large-messages.example">
<title>Large message example</title>
<para>Please see <xref linkend="examples.large-message"/> for an example which shows how
large message is configured and used with JMS.</para>
</section>
</chapter>