Re: Pro C bulk insert error
Date: Mon, 11 Oct 2010 12:51:41 +0100
All I can suggest is that when the bulk inserts run, some kind of memory over write is happening, that trashes something to do with the Oracle connection. Are you able to run the code through a debugger that checks memory access or use a tool like Purify to do the same? Your description has all of the hall marks of code writing to memory locations it should not, that wil not appear when tracing is turned on because that has the side effect of padding memory with zeros and stopping the problem from occurring.
On Fri, Oct 8, 2010 at 7:36 PM, Lou Avrami <avramil_at_concentric.net> wrote:
> Hi Pete,
> Thanks for the reply.
> The database and the application are runing on the same server.
> SMON is not producing any trace files related to the connection drop.
> the funny thing about the ORA-12152 errors are that they seem to occur only when these bulk inserts happen. When the app starts up, it makes a persistent connection to the database. Data gets loaded almost constantly under normal circumstances. It's only when there is a "site storm" that messages can back up and get queued. It's when the app tries to bulk insert many messages that the ORA-12152 error occurs.
> It's almost as though DCD, Dead Connection Detection or IDLE_TIMEOUT is enabled - the connection goes dead, then the next SQL statement fails. I am absolutely sure that DCD is not enabled ...