Archives of the TeradataForum

Message Posted: Thu, 13 Nov 2003 @ 22:16:18 GMT

	<Prev	Next>		<<First	<Prev	Next>	Last>>

I've seen the same in Type-3 driver Connection pools, in 2.2.1.7 versions. But I think its fixed from 3.0.0.1 onwards ( read this in a DR report).

This occurs when socket at one end is being killed abrubtly. You can replicate this by running a Java program, ( add a sleep statement or read statement to wait for your input ) and then do a Ctrl+C, this way socket connection will be terminated resulting the same error. It could be of 3 reasons.

1. A TPA reset that's occuring or

2. Teradata RDBMS setting which kills the stale session's after some days. Interestingly instead doing an ABORT the system might be sending a SIGKILL or SIGTERM signal causing the socket on the server end to fail.

3. Teradata Manager property that's set for aborting stale sessions.

or in worst case the fix applied in new versions is not working!!!!! ( Ask NCR ).

If you are using Weblogic/Websphere connection pooling there's something called "Refresh Interval" . Try implementing the same functionality in your pool if you don't have one. Its nothing but firing a query to the DBMS in frequent intervals and making sure that connections are working.

FYI, Pipe is a Kernel Buffer in Unix. Its a common error in network communications thrown during testing any C-Network programs. Pipe has a read end and write end -Broken Pipe is generated when READ end expects a specific number of bytes but WRITE supplies with less or vice versa, thus not satisfying the protocol defined for the communication.

Hope this helps....

Regards,

Raj

	<Prev	Next>		<<First	<Prev	Next>	Last>>