db-derby-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Brett Bergquist (Updated) (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (DERBY-5560) Java deadlock between LogicalConnection40 and ClientXAConnection40
Date Tue, 03 Jan 2012 15:18:39 GMT

     [ https://issues.apache.org/jira/browse/DERBY-5560?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Brett Bergquist updated DERBY-5560:
-----------------------------------

    Attachment: DERBY-5560.patch

After looking at the code and locking patterns between the LogicalConnection and ClientPooledConnection,
I have determined that the safest change is to just change the LogicalConnection.close to
first synchronize on the ClientPooledConnection.

The ClientPooledConnection first synchronizes on itself and then calls the LogicalConnection.nullPhysicalConnection
which synchronizes on the LogicalConnection.  So that lock pattern is ClientPooledConnection
and then LogicalConnection. 

The patch changes the LogicalConnection.close method to do the same.  The patch makes the
method not synchronized and then first synchronizes on the ClientPooledConnection and then
the LogicalConnection.  The pattern is now the same.


                
> Java deadlock between LogicalConnection40 and ClientXAConnection40
> ------------------------------------------------------------------
>
>                 Key: DERBY-5560
>                 URL: https://issues.apache.org/jira/browse/DERBY-5560
>             Project: Derby
>          Issue Type: Bug
>          Components: Network Client
>    Affects Versions: 10.8.2.2
>         Environment: Solaris 10
> Glassfish V2.1.1 
> ClientXADataSource connection pool setup to close all connections on any error
>            Reporter: Brett Bergquist
>         Attachments: DERBY-5560.patch
>
>
> There is a Java deadlock between LogicalConnection40 and ClientXAConnection40.  The order
of calls that cause the deadlock are:
> Thread 1
> ----
> LogicalConnection.close
> ClientPooledConnection.recycleConnection
> Thread 2
> ----
> ClientPooledConnection.close
> LogicalConnection.nullPhysicalConnection
> Thread 1 acquires a lock on the LogicalConnection and attempts to acquire a lock on the
ClientPooledConnection
> Thread 2 acquires a lock on the ClientPooledConnection and attempts to acquire a lock
on the LogicalConnection
> In production this occurs when one thread is committing a transaction and another thread
is trying to close the connection.  This occurred because the Glassfish connection pool is
setup to close all connections on any error on any connection and an error has been detected
on another connection in the pool.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira

        

Mime
View raw message