Bug 779781 (SOA-2143) - Active JMX connection causes deadlock/race in jbrules.esb startup
Summary: Active JMX connection causes deadlock/race in jbrules.esb startup
Keywords:
Status: CLOSED NOTABUG
Alias: SOA-2143
Product: JBoss Enterprise SOA Platform 5
Classification: JBoss
Component: JBoss Rules, Monitoring and Management
Version: 5.0.0 GA
Hardware: Unspecified
OS: Unspecified
high
high
Target Milestone: ---
: ---
Assignee: Default User
QA Contact:
URL: http://jira.jboss.org/jira/browse/SOA...
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2010-07-05 08:39 UTC by Jan Ypma
Modified: 2010-07-05 12:00 UTC (History)
0 users

Fixed In Version:
Doc Type: Bug Fix
Doc Text:
Clone Of:
Environment:
Fedora 13 x86_64, OpenJDK 1.6.0_18
Last Closed: 2010-07-05 12:00:28 UTC
Type: Bug


Attachments (Terms of Use)
JBossEAPTimeout.java (3.90 KB, text/plain)
2010-07-05 08:40 UTC, Jan Ypma
no flags Details


Links
System ID Private Priority Status Summary Last Updated
Red Hat Issue Tracker SOA-2143 0 None None None Never

Description Jan Ypma 2010-07-05 08:39:16 UTC
Steps to Reproduce: - Install clean deployment of JBoss SOA-P 5.0.1
- Disable JMX authentication [1]
- Run the attached unit test.

[1] http://www.redhat.com/docs/en-US/JBoss_Enterprise_Application_Platform/5.0.1/html/Installation_Guide/Disabling_Authentication.html
project_key: SOA

Originally found that maven-jboss-plugin doesn't detect JBoss SOA-P startup correctly. I created a unit test that reproduces the behaviour. Basically, the plugin starts JBoss, creates JMX connection, and polls "jboss.system:type=Server" for the "Started" attribute to become true.

When launching JBoss this way, it'll start deploying services fine until (just before or after) jbrules.esb. Just before logging deployment start of jbrules.esb, the JBoss process halts (pauses) completely. After the unit test times out, and the JMX connection is broken, JBoss resumes startup, and finishes startup correctly. Forcibly terminating the unit test earlier has JBoss resume earlier as well. The JMX poll calls themselves are not blocking in the unit test. So it's the active JMX connection itself that causes JBoss SOA-P to hang around somewhere near deployment of jbrules.esb.

Comment 1 Jan Ypma 2010-07-05 08:40:04 UTC
Test case that reproduces the defect

Comment 2 Jan Ypma 2010-07-05 08:40:04 UTC
Attachment: Added: JBossEAPTimeout.java


Comment 3 Jan Ypma 2010-07-05 12:00:28 UTC
Not defect in jboss, but in maven-jboss-plugin. Consule buffer isn't periodically flushed on linux, causing jboss to halt after a certain amount of output is generated and a buffer is full (probably ??kbytes).


Note You need to log in before you can comment on or make changes to this bug.