[jboss-jira] [JBoss JIRA] Created: (JBREM-762) Deadlock bug in socket invoker
Tim McCune (JIRA)
jira-events at lists.jboss.org
Thu Jun 28 17:56:57 EDT 2007
Deadlock bug in socket invoker
------------------------------
Key: JBREM-762
URL: http://jira.jboss.com/jira/browse/JBREM-762
Project: JBoss Remoting
Issue Type: Bug
Security Level: Public (Everyone can see)
Components: transport
Affects Versions: 1.4.6.GA
Environment: JBoss 4.0.4 EJB 3.0-rc9-patch1 JDK 1.6.0
Reporter: Tim McCune
There is a potential thread leak in ServerThread.java that causes ServerThreads to stick around, but to disappear from the thread pool as the accept() thread(s) see it. Eventually, when the # of leaked threads reaches the MaxPoolSize, the system ends up deadlocked and will not accept any more incoming socket connections. I have described the problem in great detail in a forum post and in support case 00016786. I have patched the code, and so far it seems to be working fine and to have fixed the problem. This is what I patched:
In ServerThread.java, I wrapped the contents of wakeup() inside of a try/catch, block, with the following code in the catch block:
} catch (Exception e) {
synchronized (clientpool) {
synchronized (threadpool) {
clientpool.remove(this);
threadpool.add(this);
Thread.interrupted();
}
}
throw e;
}
This makes sure that the thread is put back into the threadpool when an exception occurs in the createServerSocket() method (this method is doing socket I/O in 1.4.6, and in our case, this socket connection was timing out.).
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators: http://jira.jboss.com/jira/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
More information about the jboss-jira
mailing list