服务器闲置几个小时后，ActiveMQ停止接收消息

Question

I've been browsing the forums for last few days and tried almost everything i could find, but without any luck.

The situation is: inside our Java Web Application we have ActiveMQ 5.7 (I know it's very old, eventually we will upgrade to newer version - but for some reasons it's not possible right now). We have only one broker and multiple consumers.

When I start the servers (I have tried to do so for 2, 3, 4 and more servers) everything is ok. The servers are comunicating with each other, QUEUE messages are consumed instantly. But when I leave the servers idle (for example to finally catch some sleep ;) ) it is no longer the case. Messages are stuck in the database and are not beign consumed. The only option to have them delivered is to restart the server.

Part of my configuration (we keep it in properties file, it's the actual state, however I have tried many different combinations):

BrokerServiceURI=broker:(tcp://0.0.0.0:{0})/{1}?persistent=true&useJmx=false&populateJMSXUserID=false&useShutdownHook=false&deleteAllMessagesOnStartup=false&enableStatistics=true 
ConnectionFactoryURI=failover://({0})?initialReconnectDelay=100&timeout=6000 
ConnectionFactoryServerURI=tcp://{0}:{1}?keepAlive=true&soTimeout=100&wireFormat.cacheEnabled=false&wireFormat.tightEncodingEnabled=false&wireFormat.maxInactivityDuration=0 
BrokerService.startAsync=true 
BrokerService.networkConnectorStartAsync=true 
BrokerService.keepDurableSubsActive=false

Do you have a clue?

Answer 1

I cannot actually tell you the reason from the description mentioned above but I can list down a few checks that are fresh in my mind. Please confirm the following if they are valid for you or not.

1. Can you check the consumer connections?
2. Are the consumer sessions still active?
3. If all the consumer-connections are up, then check the thread-dump whether the active consumer threads (I'm assuming you created consumer threads, correct me if I'm wrong) are in RUNNING or WAITING state(this happened with me where all the consumers were active but some other thread was keeping a lock on Logger while posting a message to slack and the consumers were in WAITING state) because of some other thread in the server).
4. Check the Dispatch queue size for each consumer. Check the prefetch of each consumer and then compare Dispatch Queue size with Prefetch, refer
5. Is there a JMSXGroupID you are allotting to each message?

Can you tell a little more about your consumer/producer/broker configurations?