[AMQ-4106] STOMP+NIO and Heartbeats gives consistent InactivityIOException

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 5.7.0
• Fix Version/s: 5.8.0
• Component/s: STOMP

Environment

Kernel:Linux tjjackson 3.0.0-26-generic #43-Ubuntu SMP Tue Sep 25 17:19:22 UTC 2012 x86_64 x86_64 x86_64 GNU/Linux

AMQ: 5.7.0
Ubuntu: 11.10
Java: 1.6.0_24

From activemq.xml:

<transportConnector name="stomp"
uri="stomp+nio://0.0.0.0:61613?trace=true"/>

• Assignee: Timothy A. Bish
• Reporter: Guy Allard
• Created: 14/Oct/12 14:22
• Updated: 23/Oct/12 19:16
• Resolved: 23/Oct/12 19:14

Description

When a stomp 1.1 client:

a) connects to a stomp+nio transport
b) elects to send heartbeats when the network is idle

the first several heartbeat sends appear to succeed, but then the connection is forcibly closed by the broker.

From activemq.log:

2012-10-14 09:53:07,755 | WARN | Transport Connection to: tcp://127.0.0.1:36440 failed: org.apache.activemq.transport.InactivityIOException: Channel was inactive for too (>10000) long: tcp://127.0.0.1:36440 | org.apache.activemq.broker.TransportConnection.Transport | ActiveMQ InactivityMonitor Worker

From wrapper.log:

INFO | jvm 1 | 2012/10/14 09:53:07 | WARN | Transport Connection to: tcp://127.0.0.1:36440 failed: org.apache.activemq.transport.InactivityIOException: Channel was inactive for too (>10000) long: tcp://127.0.0.1:36440

I am able to reproduce this with either:

a) the current Ruby stomp gem
b) the current golang stompngo package

It is unclear to me whether:

a) the AMQ heartbeat code is not tolerant as the specification suggests
b) the AMQ heartbeat receiver is not reading or recognizing heartbeats at all

In all cases these tests succeed with either:

a) AMQ 5.7.0 stomp without nio
b) Apollo
c) RabbitMQ
d) a client that only receives heartbeats, but does not send them

I will attach a Ruby client example shortly.