WebLogic is not responding to client request under heavy load

Question

Under heavy loading, our WLS 6.0 SP2 will be not be responding to client requests
after some time, say, 1 hour. From the administration console, the request throughput
is 0 while request waiting accumulates to over 1,000. Even the clients stop making
request to the server, the request waiting will not drop. It seems that WLS blocked
on some threads and unable to serve the requests (see attached thread dump for
details).
Our configuration is as follow:
OS: SunOS 5.7
Database: MS SQL Server 7.0 running on NT4 Enterprise Edition.
(JDBC Driver is the one bundled with WLS 6.0)
We know that using 'DriverManager.getConnection()' can cause deadlock in number
of cases but we have gone through all the code but all of our Database connections
are obtained through datasource. And the connection is closed properly in finally
block.
Moreover, the application is originally running on WLS 4.51 but no problem is
encountered.
Does any expert know what the problem and solution are?
Thanks!
[session.log]

Answer 1

It appears that some threads were possibly blocked at some jDriver calls
initially (ExecuteThread 0,10,15,23 of the default thread pool). WLS JTA
subsequently timed out and rolled back the transactions asynchronously. The
first rollback attempt was blocked at the jDriver level (ExecuteThread 18 of
the default thread pool). Each subsequent rollback retry blocks an
additional execute thread (due to a JTA bug that is fixed in WLS 6.1, but
not 6.0 SPs) - ExecuteThread 1-9,11-26 of the default thread pool.
Eventually, the server ran out of execute threads and became unresponsive.
It is unclear that whether the initial blocking of threads by the jDriver is
a jDriver issue or an application issue. Please report to BEA support at
[email protected] for further assistance.
Regards,
Priscilla
Gary Mok <[email protected]> wrote in message
news:[email protected]..
>
Under heavy loading, our WLS 6.0 SP2 will be not be responding to clientrequests
after some time, say, 1 hour. From the administration console, the requestthroughput
is 0 while request waiting accumulates to over 1,000. Even the clientsstop making
request to the server, the request waiting will not drop. It seems thatWLS blocked
on some threads and unable to serve the requests (see attached thread dumpfor
details).
Our configuration is as follow:
OS: SunOS 5.7
Database: MS SQL Server 7.0 running on NT4 Enterprise Edition.
(JDBC Driver is the one bundled with WLS 6.0)
We know that using 'DriverManager.getConnection()' can cause deadlock innumber
of cases but we have gone through all the code but all of our Databaseconnections
are obtained through datasource. And the connection is closed properly infinally
block.
Moreover, the application is originally running on WLS 4.51 but no problemis
encountered.
Does any expert know what the problem and solution are?
Thanks!