Spring integration - Queue/Poller seems to exhaust threadpool without any action

Question

I have a Spring integration app, attached to an AMQP broker.

I want to receive messages from an amqp-queue, and update db records.

In order to improve performance, I have a pool of workers allowing multiple updates to occur concurrently.

I have the following configuration:

<int-amqp:inbound-channel-adapter queue-names="pricehub.fixtures.priceUpdates.queue" 
                            channel="pricehub.fixtures.priceUpdates.channel"
                            message-converter="jsonMessageConverter"/>

<int:channel id="pricehub.fixtures.priceUpdates.channel">
    <int:queue  />
</int:channel>

<int:service-activator ref="updatePriceAction" 
     method="updatePrices" 
     input-channel="pricehub.instruments.priceUpdates.channel">
    <int:poller fixed-delay="50" time-unit="MILLISECONDS" task-executor="taskExecutor" />
</int:service-activator>

<task:executor id="taskExecutor" pool-size="5-50" keep-alive="120" queue-capacity="500"/>

If I start this running, with no inbound messages to process on the AMQP channel, I quickly see the thredpool get exhausted, and start rejecting.

Here's the logs:

[Thu Apr 2013 23:41:51.153] DEBUG [] (org.springframework.amqp.rabbit.listener.BlockingQueueConsumer:185) - Retrieving delivery for Consumer: tag=[amq.ctag-w4qPp60jVEQOIEovR4cERv], channel=Cached Rabbit Channel: AMQChannel(amqp://guest@127.0.0.1:5672/,1), acknowledgeMode=AUTO local queue size=0
[Thu Apr 2013 23:41:51.160] DEBUG [] (org.springframework.amqp.rabbit.listener.BlockingQueueConsumer:185) - Retrieving delivery for Consumer: tag=[amq.ctag-Q3Lq4R9g9E8WBNVLYzaFmq], channel=Cached Rabbit Channel: AMQChannel(amqp://guest@127.0.0.1:5672/,2), acknowledgeMode=AUTO local queue size=0
[Thu Apr 2013 23:41:51.166] DEBUG [] (org.springframework.amqp.rabbit.listener.BlockingQueueConsumer:185) - Retrieving delivery for Consumer: tag=[amq.ctag-w8bg7ltEV2mot8QXDPCmfK], channel=Cached Rabbit Channel: AMQChannel(amqp://guest@127.0.0.1:5672/,3), acknowledgeMode=AUTO local queue size=0
[Thu Apr 2013 23:41:51.170] DEBUG [] (org.springframework.amqp.rabbit.listener.BlockingQueueConsumer:185) - Retrieving delivery for Consumer: tag=[amq.ctag-A-0KdqhFjpc-Hvjmv7aZAc], channel=Cached Rabbit Channel: AMQChannel(amqp://guest@127.0.0.1:5672/,4), acknowledgeMode=AUTO local queue size=0
[Thu Apr 2013 23:41:51.180] DEBUG [] (org.springframework.integration.endpoint.PollingConsumer:71) - Received no Message during the poll, returning 'false'
[Thu Apr 2013 23:41:51.180] DEBUG [] (org.springframework.integration.endpoint.PollingConsumer:71) - Received no Message during the poll, returning 'false'
[Thu Apr 2013 23:41:51.199] DEBUG [] (org.springframework.integration.endpoint.PollingConsumer:71) - Received no Message during the poll, returning 'false'
[Thu Apr 2013 23:41:51.200] DEBUG [] (org.springframework.integration.endpoint.PollingConsumer:71) - Received no Message during the poll, returning 'false'
[Thu Apr 2013 23:41:51.220] DEBUG [] (org.springframework.integration.endpoint.PollingConsumer:71) - Received no Message during the poll, returning 'false'

Pretty quickly, the thread pool starts rejecting the executions:

[Thu Apr 2013 23:47:15.363] ERROR [] (org.springframework.integration.handler.LoggingHandler:126) - org.springframework.core.task.TaskRejectedException: Executor [java.util.concurrent.ThreadPoolExecutor@6ff3cb0e] did not accept task: org.springframework.integration.util.ErrorHandlingTaskExecutor$1@78615c8b
    at org.springframework.scheduling.concurrent.ThreadPoolTaskExecutor.execute(ThreadPoolTaskExecutor.java:244)
    at org.springframework.integration.util.ErrorHandlingTaskExecutor.execute(ErrorHandlingTaskExecutor.java:49)
    at org.springframework.integration.endpoint.AbstractPollingEndpoint$Poller.run(AbstractPollingEndpoint.java:231)
    at org.springframework.scheduling.support.DelegatingErrorHandlingRunnable.run(DelegatingErrorHandlingRunnable.java:53)
    at org.springframework.scheduling.concurrent.ReschedulingRunnable.run(ReschedulingRunnable.java:81)
    at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:439)
    at java.util.concurrent.FutureTask$Sync.innerRun(FutureTask.java:303)
    at java.util.concurrent.FutureTask.run(FutureTask.java:138)
    at java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask.access$301(ScheduledThreadPoolExecutor.java:98)
    at java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask.run(ScheduledThreadPoolExecutor.java:206)
    at java.util.concurrent.ThreadPoolExecutor$Worker.runTask(ThreadPoolExecutor.java:886)
    at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:908)
    at java.lang.Thread.run(Thread.java:680)
Caused by: java.util.concurrent.RejectedExecutionException
    at java.util.concurrent.ThreadPoolExecutor$AbortPolicy.rejectedExecution(ThreadPoolExecutor.java:1768)
    at java.util.concurrent.ThreadPoolExecutor.reject(ThreadPoolExecutor.java:767)
    at java.util.concurrent.ThreadPoolExecutor.execute(ThreadPoolExecutor.java:658)
    at org.springframework.sched

uling.concurrent.ThreadPoolTaskExecutor.execute(ThreadPoolTaskExecutor.java:241) ... 12 more

I suspect that the culprit lies here: BlockingQueueConsumer - indicating that each poll for a message blocks the thread until a message arrives ... leading to the threadpool being exhausted quickly.

What's the correct way to configure this?

Answer 1

Rather than using a QueueChannel and poller, why not simply increase the concurrent-consumers attribute on the inbound adapter?

    <xsd:attribute name="concurrent-consumers" type="xsd:string">
        <xsd:annotation>
            <xsd:documentation>
Specify the number of concurrent consumers to create. Default is 1.
Raising the number of concurrent consumers is recommended in order to scale the consumption of messages coming in
from a queue. However, note that any ordering guarantees are lost once multiple consumers are registered. In
general, stick with 1 consumer for low-volume queues.
            </xsd:documentation>
        </xsd:annotation>
    </xsd:attribute>

And, remove the <queue/> and <poller/>.

Also, I always recommend including the thread name in the log (%t for log4J); it makes it easier to debug threading issues.

EDIT:

With the poller, the reason you are running out of threads is the poller has a default receive-timeout of 1 second. You are scheduling a thread every 50ms, but each then waits in the QueueChannel for 1 second. Eventually your task queue fills up.

To avoid this, simply set the receive-timeout to 0 on the <poller/> if you wish to continue with this technique - but using higher concurrency in the adapter is more efficient because there's no polling or handover to another thread.

Answer 2

It appears I needed a bridge to map between the amqp-inbound queue (which is a pub/sub style queue), and a queue-channel.

<int-amqp:inbound-channel-adapter queue-names="pricehub.fixtures.priceUpdates.queue" 
                            channel="pricehub.fixtures.priceUpdates.subpub"
                            message-converter="jsonMessageConverter"/>

<int:publish-subscribe-channel id="pricehub.fixtures.priceUpdates.subpub" />
<int:bridge input-channel="pricehub.fixtures.priceUpdates.subpub" 
    output-channel="pricehub.fixtures.priceUpdates.channel" />

<int:channel id="pricehub.fixtures.priceUpdates.channel">
    <int:queue  />
</int:channel>

<int:service-activator ref="updatePriceAction" 
     method="updatePrices" 
     input-channel="pricehub.instruments.priceUpdates.channel">
    <int:poller fixed-delay="50" time-unit="MILLISECONDS" task-executor="taskExecutor" />
</int:service-activator>

<task:executor id="taskExecutor" pool-size="5-50" keep-alive="120" queue-capacity="500"/>

This seems like a LOT of code to achieve a fairly trivial task - so if anyone has better solutions, or suggestions for improvements, I'd love to see them.