Unexpected end of QFJ Timer Thread

[cd-ppi]

We have an application that is connected to several other sites to receive FIX messages. Works fine most of the time, but our customer has noticed at least two instances where connectivity was lost without any notice - no exception, no error logged. The most obvious symptom is that the QFJ Timer Thead stops to work. One moment it is happily reporting outgoing messages for the heartbeat, and then it's gone. There is also no sign of it at the planned logout time.

At least in the last instance (mayb in both), the session that was last reported for an outgoing heartbeat message is also gone. there are no more incoming messages reported for this session, although the counterpart has said that they've sent two more messages a couple of hours later.

I'm aware that this is hardly enough information. I would appreciate any pointers on how to tackle this issue.

Many thanks!

system information:

• OS: Windows,
• Java 17
• QFJ 2.3.1

[chrjohn]

Just to clarify: you have one application with several sessions but only two of them stopped working?

Also it is strange that the counterparty sent messages hours later. In the meantime the session should have been disconnected due to a heartbeat timeout.

My first suspicion would be that your application hangs somewhere in a fromApp() or fromAdmin() call, but anyway: the next time this happens you should take a thread dump of the application. That way you can see if the QFJ Timer Thread really died or if it is blocked in some other thread.

[cd-ppi]

It is indeed one application with several sessions. After the heartbeat stops, we still see incoming messages for several sessions, but not for the one that reported the last outgoing heartbeat. And for that session, our customer reported that expected messages did not arrive. They say that the sender said that they delivered two messages, but there might be some confusion involved.

So we assume that one session is down or blocked. This has happened on two occasions, with different counterparts.

I've instructed our customer on how to obtain a thread dump.

Thanks for the help!