Apache Artemis跨版本联邦环境消费大消息遇NegativeArraySizeException

问题描述 #

使用ActiveMQ Artemis Core Jakarta Messaging 2.37.0消费消息时，抛出NegativeArraySizeException，控制台标记该消息为大消息，无法直接获取负载。环境包含3个运行2.36.0版本的broker节点，客户端使用2.37.0版本。消息在节点间流转，消息数波动，消费者持续收到消息但无法读取，且消息不会进入死信队列（DLQ）。需要解决如何恢复或检查消息负载的问题。

错误堆栈 #

2025-04-22 17:31:50.545  WARN [topic#eu.JmsListenerAnnotationBeanPostProcessor#1-12] - (eu.JmsListenerAnnotationBeanPostProcessor$1$1:1092) - Setup of JMS message listener invoker failed for destination 'example.user.v1.roles' - trying to recover. Cause: -1663738403
java.lang.NegativeArraySizeException: -1663738403
 at org.apache.activemq.artemis.api.core.SimpleString.readSimpleString(SimpleString.java:257) ~[artemis-commons-2.37.0.jar:2.37.0]
 at org.apache.activemq.artemis.api.core.SimpleString.readSimpleString(SimpleString.java:243) ~[artemis-commons-2.37.0.jar:2.37.0]
 at org.apache.activemq.artemis.api.core.SimpleString.readNullableSimpleString(SimpleString.java:230) ~[artemis-commons-2.37.0.jar:2.37.0]
 at org.apache.activemq.artemis.core.buffers.impl.ChannelBufferWrapper.readNullableSimpleString(ChannelBufferWrapper.java:79) ~[artemis-commons-2.37.0.jar:2.37.0]
 at org.apache.activemq.artemis.reader.TextMessageUtil.readBodyText(TextMessageUtil.java:37) ~[artemis-core-client-2.37.0.jar:2.37.0]
 at org.apache.activemq.artemis.jms.client.ActiveMQTextMessage.doBeforeReceive(ActiveMQTextMessage.java:106) ~[artemis-jakarta-client-2.37.0.jar:2.37.0]
 at org.apache.activemq.artemis.jms.client.ActiveMQMessageConsumer.getMessage(ActiveMQMessageConsumer.java:230) ~[artemis-jakarta-client-2.37.0.jar:2.37.0]
 at org.apache.activemq.artemis.jms.client.ActiveMQMessageConsumer.receive(ActiveMQMessageConsumer.java:134) ~[artemis-jakarta-client-2.37.0.jar:2.37.0]
 at org.springframework.jms.support.destination.JmsDestinationAccessor.receiveFromConsumer(JmsDestinationAccessor.java:132) ~[spring-jms-6.2.1.jar:6.2.1]
 at org.springframework.jms.listener.AbstractPollingMessageListenerContainer.receiveMessage(AbstractPollingMessageListenerContainer.java:441) ~[spring-jms-6.2.1.jar:6.2.1]

更新记录 #

更新1 - 2025/04/29 #

切换到2.36.0版本后问题依旧，执行artemis browser命令报错：

$ artemis browser --user artemis --password artemis --url 'tcp://messaging-01.dc.eu:61616?sslEnabled=true' --destination domain-events.#::audit-trail_shared
Connection brokerURL = tcp://messaging-01.dc.eu:61616?sslEnabled=true
Consumer:: filter = null
Consumer ActiveMQQueue[domain-events.#::audit-trail_shared], thread=0 trying to browse 1000 messages
2025-04-29 10:54:07,448 ERROR [org.apache.activemq.artemis.jms.client] AMQ134002: Queue Browser failed to create message ClientLargeMessageImpl[messageID=956935038, durable=true, address=domain-events.global.customer-service.v1.customers,userID=63df8053-20d5-11f0-84de-9a2c728d6a7d,properties=TypedProperties[ResourceType=CUSTOMER, EventVersion=v1, _AMQ_Hops=[0000 0001], Address=domain-events.global.customer-service.v1.customers, Authorization={"jwtPrincipal":{"tokenValue":"eyJhbGQy...c53835cbd"}},"authorities":[]}, _AMQ_ROUTING_TYPE=0, AuthorizationType=jwt-v1.0, _AMQ_LARGE_COMPRESSED=false, OperationType=CREATE, EventDateTime=2025-04-24T06:28:52.318060801Z, _AMQ_LARGE_SIZE=130153, __AMQ_CID=63d7b820-20d5-11f0-84de-9a2c728d6a7d, SendOffsetDateTime=2025-04-24T06:28:52.318060801Z, timestamp=1745476132318]]
java.lang.NegativeArraySizeException: -1662157365
        at org.apache.activemq.artemis.api.core.SimpleString.readSimpleString(SimpleString.java:257) ~[artemis-commons-2.36.0.jar:2.36.0]
        at org.apache.activemq.artemis.api.core.SimpleString.readSimpleString(SimpleString.java:243) ~[artemis-commons-2.36.0.jar:2.36.0]
        at org.apache.activemq.artemis.api.core.SimpleString.readNullableSimpleString(SimpleString.java:230) ~[artemis-commons-2.36.0.jar:2.36.0]
        at org.apache.activemq.artemis.core.buffers.impl.ChannelBufferWrapper.readNullableSimpleString(ChannelBufferWrapper.java:79) ~[artemis-commons-2.36.0.jar:2.36.0]
        at org.apache.activemq.artemis.reader.TextMessageUtil.readBodyText(TextMessageUtil.java:37) ~[artemis-core-client-2.36.0.jar:2.36.0]
        at org.apache.activemq.artemis.jms.client.ActiveMQTextMessage.doBeforeReceive(ActiveMQTextMessage.java:106) ~[artemis-jms-client-2.36.0.jar:2.36.0]
        at org.apache.activemq.artemis.jms.client.ActiveMQQueueBrowser$BrowserEnumeration.nextElement(ActiveMQQueueBrowser.java:139) [artemis-jms-client-2.36.0.jar:2.36.0]
        at org.apache.activemq.artemis.jms.client.ActiveMQQueueBrowser$BrowserEnumeration.nextElement(ActiveMQQueueBrowser.java:110) [artemis-jms-client-2.36.0.jar:2.36.0]
        at org.apache.activemq.artemis.cli.commands.messages.ConsumerThread.browse(ConsumerThread.java:127) [artemis-cli-2.36.0.jar:2.36.0]
        at org.apache.activemq.artemis.cli.commands.messages.ConsumerThread.run(ConsumerThread.java:69) [artemis-cli-2.36.0.jar:2.36.0]
Consumer ActiveMQQueue[domain-events.#::audit-trail_shared], thread=0 browsed: 0 messages
Consumer ActiveMQQueue[domain-events.#::audit-trail_shared], thread=0 Browser thread finished

更新2 - 2025/05/02 #

生产者始终通过URL配置:61616?compressLargeMessages=true启用压缩发送消息。手动解压/data/large-message文件夹中的消息负载：

cat 872923939.msg | python3 -c "import sys, zlib; print(zlib.decompress(sys.stdin.buffer.read()).decode('latin'))"

输出开头有两个意外字符（如½¬），怀疑影响客户端解压。近期将Artemis客户端版本从2.31.x升级到2.37.x。

更新3 - 2025/05/28 #

发现消息在federation-A生产，流转到federation-B后，federation-B的消费者接收时触发异常。federation-A的broker.xml配置：

<connectors>
    <connector name="artemis">tcp://qa-messaging-01.example.com:61616</connector>
    <connector name="cloud-01-connector">tcp://qa-messaging-federation-01.example.com:61616?sslEnabled=true;ackBatchSize=100;consumerWindowSize=-1</connector>
    <connector name="clous-02-connector">tcp://qa-messaging-federation-02.example.com:61616?sslEnabled=true;ackBatchSize=100;consumerWindowSize=-1</connector>
    <connector name="cloud-03-connector">tcp://qa-messaging-federation-03.example.com:61616?sslEnabled=true;ackBatchSize=100;consumerWindowSize=-1</connector>
</connectors>

<federations>
    <federation name="federation-A" user="federation-user" password="******">
        <upstream name="cloud-qa">
            <static-connectors>
                <connector-ref>cloud-01-connector</connector-ref>
                <connector-ref>cloud-02-connector</connector-ref>
                <connector-ref>cloud-03-connector</connector-ref>
            </static-connectors>
            <policy ref="domain-events-federation"/>
        </upstream>
        <address-policy name="domain-events-federation" max-hops="1" auto-delete="false">
            <include address-match="domain-events.example.#"/>
        </address-policy>
    </federation>
</federations>

解决方案建议 #

1. 修复消息负载的异常前缀 #

手动处理大消息文件，去掉开头的两个意外字符后重新解压，确认消息内容是否完整：

cat 872923939.msg | tail -c +3 | python3 -c "import sys, zlib; print(zlib.decompress(sys.stdin.buffer.read()).decode('utf-8'))"

如果内容正常，说明消息在传输或存储过程中被额外写入了前缀字节，导致客户端解析时错误计算数组大小。

2. 统一客户端与Broker版本 #

将Broker版本升级到2.37.0，保持客户端与Broker版本一致，验证是否因版本差异导致压缩逻辑不兼容。

3. 调整联邦连接器的压缩配置 #

在federation-A的连接器配置中添加compressLargeMessages=true，确保联邦转发时保持压缩状态，避免消息负载被篡改：

<connector name="cloud-01-connector">tcp://qa-messaging-federation-01.example.com:61616?sslEnabled=true;ackBatchSize=100;consumerWindowSize=-1;compressLargeMessages=true</connector>

同时检查federation-B的Broker是否启用压缩支持，确保两端配置一致。

4. 删除损坏消息 #

若消息无法修复，可通过Artemis CLI删除指定消息：

artemis queue remove --user artemis --password artemis --url 'tcp://messaging-01.dc.eu:61616?sslEnabled=true' --queue domain-events.#::audit-trail_shared --message-id 956935038

或清空整个队列（需确认数据可丢弃）：

artemis queue purge --user artemis --password artemis --url 'tcp://messaging-01.dc.eu:61616?sslEnabled=true' --queue domain-events.#::audit-trail_shared

5. 启用大消息校验 #

在Broker的broker.xml中添加大消息校验配置，拦截损坏消息：

<large-messages-directory>data/large-messages</large-messages-directory>
<max-large-message-size>104857600</max-large-message-size>
<enable-large-messages>true</enable-large-messages>
<validate-large-messages>true</validate-large-messages>

内容的提问来源于stack exchange，提问作者André Conrado