ActiveMQ InactivityMonitor 参数未应用于 HTTP 传输连接

Question

In my team we are using ActiveMQ 5.15.11 in production and our messages consumers are connected through HTTP protocol. Since we have network issues on the consumer side, we tried to disable the InactivityMonitor in the activemq.xml (see below)

    <transportConnectors>
        <transportConnector name="openwire" uri="tcp://host:61616?maximumConnections=1000&amp;wireFormat.maxFrameSize=104857600"/>
        <transportConnector name="http" uri="http://host:61617?wireFormat.maxInactivityDuration=0"/>
    </transportConnectors>

It seems like the " wireFormat.maxInactivityDuration=0 "is not applied since we have logs from InactivityMonitor in the consumer :

2020-03-18 00:00:03,426 [DEBUG] -  -  - WriteChecker: 10000ms elapsed since last write check.
2020-03-18 00:00:03,426 [DEBUG] -  -  - Running WriteCheck[https://host:443/activemq]
2020-03-18 00:00:03,509 [DEBUG] -  -  - WriteChecker: 10000ms elapsed since last write check.
2020-03-18 00:00:03,509 [DEBUG] -  -  - Running WriteCheck[https://host:443/activemq]
2020-03-18 00:00:03,514 [DEBUG] -  -  - WriteChecker: 10000ms elapsed since last write check.
2020-03-18 00:00:03,514 [DEBUG] -  -  - Running WriteCheck[https://host:443/activemq]
2020-03-18 00:00:03,642 [DEBUG] -  -  - WriteChecker: 10000ms elapsed since last write check.
2020-03-18 00:00:03,642 [DEBUG] -  -  - Running WriteCheck[https://host:443/activemq]
2020-03-18 00:00:03,706 [DEBUG] -  -  - WriteChecker: 10000ms elapsed since last write check.
2020-03-18 00:00:03,706 [DEBUG] -  -  - Running WriteCheck[https://host:443/activemq]
2020-03-18 00:00:03,738 [DEBUG] -  -  - WriteChecker: 10000ms elapsed since last write check.
2020-03-18 00:00:03,738 [DEBUG] -  -  - Running WriteCheck[https://host:443/activemq]

Furthermore, as soon as we get long network delay, we also get log :

2020-03-01 20:43:17,578 [WARN ] -  -  - Transport (https://host:443/activemq) failed , attempting to automatically reconnect: {}
org.apache.activemq.transport.InactivityIOException: Channel was inactive for too (>30000) long: https://host:443/activemq
    at org.apache.activemq.transport.AbstractInactivityMonitor$5.run(AbstractInactivityMonitor.java:246)
    at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
    at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
    at java.lang.Thread.run(Thread.java:748)

Notice 1 : ActiveMQ clients are JMS Consumers .

Notice 2 : we tried to disable InactivityMonitor on TCP transport and it works well with this protocol.

Thanks for your help in advance.

Answer 1

For the HTTP transport I don't think that the standard OpenWire protocol negotiation step is performed which likely means that the server side configuration of max-inactivity-duration is visible at the client side. This results in the client using the defaults for the inactivity checks which I recall as being 30 seconds or so which it then splits into smaller chunks for read and write checks.

You would likely need to configure the client end to not do the checking using the same wireformat options. I don't think any of that is extensively tested though so there could likely be issues with altering defaults over the HTTP transport wrappers. If it continues to misbehave then likely you'd need to open a JIRA with the ActiveMQ project to get something implemented to let you control it better.

On the client side you should be able to disable the inactivity monitor with transport.useInactivityMonitor=false or maybe just useInactivityMonitor=false on the URI.